Adding support for Dell EqualLogic PS-Series storage arrays to version 3.5.1 of the popular, open source switch and router configuration management tool RANCID.

For the impatient and TL;DR here are the extensions to RANCID for the management of Dell EqualLogic PS-Series storage arrays:

Login script for Dell EqualLogic PS-Series storage arrays
Perl module to generate, process and save the configuration of Dell EqualLogic PS-Series storage arrays

The sources are to be found in my RANCID repository on GitHub

RANCID has, in its current version 3.5.1, support for a large variety of network devices like routers, switches, load-balancers, etc. Unfortunately there is currently little or no support for the management of storage devices, even though a lot of them offer a command line interface which can be used by RANCID.

Although there probably are a couple of reasons for this, i suppose this is largely due to the fact that network and storage admins are – in most organizations – still in different groups, each with their own set of management and support tools. With RANCID originating from the realm of network administration, probably only few storage admins know about this very valuable tool to begin with. There is probably also very little transfer over from the position of network administrator into the area of storage administration and thus a limited amount of knowledge transfer between those two fields.

This blog post describes how to extend and configure RANCID in order to add support for Dell EqualLogic PS-Series storage arrays. The extensions are based on the – at the time of writing – current version 3.5.1 of RANCID. RANCID can either be build from source or be installed pre-packaged e.g. from the backports repository of Debian stable (jessie). Basically, the extension to RANCID consist of only two files:

• An expect login script for Dell EqualLogic PS-Series storage arrays (eqllogin in my RANCID repository on GitHub). This script provides the basic access to the command line interface of a EqualLogic system. It can be used for interactive sessions as well as scripted access. It is used by the following equallogic.pm RANCID perl module.

• A perl module to generate, process and save the configuration of Dell EqualLogic PS-Series storage arrays (equallogic.pm in my RANCID repository on GitHub), with the use of the aforementioned expect login script. The save-config -verbose command is used on the Dell EqualLogic PS-Series storage arrays in order to generate and print the array configuration. The processing part currently just removes some informational messages from the output of the save-config -verbose command. If the environment variable FILTER_PWDS is set in rancid.conf, the iSCSI CHAP users as well as the iSCSI target authorization usernames and passwords will be removed before saving the configuration. If the environment variable NOCOMMSTR is set in rancid.conf, the SNMP Community strings will be removed before saving the configuration.

Besides those two extensions only a small change to /etc/rancid/rancid.types.conf, as well as the standard RANCID configuration of a new device group and device is necessary. See the following full step-by-step configuration example for Dell EqualLogic PS-Series storage arrays:

• Add the backports repository of Debian stable (jessie) to the APT configuration:

  root@host:~$ echo 'deb http://http.debian.net/debian jessie-backports main non-free contrib' >> /etc/apt/sources.list.d/jessie-backports.list
  root@host:~$ apt-get update

• Install RANCID v3.5.1 from the backports repository of Debian stable (jessie):

  root@host:~$ apt-get -y install rancid/jessie-backports
  
  Reading package lists... Done
  Building dependency tree       
  Reading state information... Done
  Selected version '3.5.1-1~bpo8+1' (Debian Backports:jessie-backports [amd64]) for 'rancid'
  The following extra packages will be installed:
    expect tcl-expect
  The following NEW packages will be installed:
    expect rancid tcl-expect
  0 upgraded, 3 newly installed, 0 to remove and 0 not upgraded.
  Need to get 511 kB of archives.
  After this operation, 2,178 kB of additional disk space will be used.
  [...]

  Optional: In case Subversion should be used as a revision control system (RCS) to store the switch configuration, install it:

  root@host:~$ apt-get -y install subversion
  
  Reading package lists... Done
  Building dependency tree       
  Reading state information... Done
  The following extra packages will be installed:
    libapr1 libaprutil1 libldap-2.4-2 libsasl2-2 libsasl2-modules libsasl2-modules-db libserf-1-1 libsvn1
  Suggested packages:
    libsasl2-modules-otp libsasl2-modules-ldap libsasl2-modules-sql libsasl2-modules-gssapi-mit libsasl2-modules-gssapi-heimdal subversion-tools db5.3-util patch
  The following NEW packages will be installed:
    libapr1 libaprutil1 libldap-2.4-2 libsasl2-2 libsasl2-modules libsasl2-modules-db libserf-1-1 libsvn1 subversion
  0 upgraded, 9 newly installed, 0 to remove and 0 not upgraded.
  Need to get 2,723 kB of archives.
  After this operation, 9,683 kB of additional disk space will be used.
  [...]

• Download the login script for Dell EqualLogic PS-Series storage arrays and store it under the path /usr/lib/rancid/bin/.

• Download the Perl module to process and save the configuration of Dell EqualLogic PS-Series storage arrays and store it under the path /usr/share/perl5/rancid/.

• Edit the global RANCID configuration:

  root@host:~$ vi /etc/rancid/rancid.conf

  Select the RCS (CVS, SVN or Git) of your choice. In this example SVN is used:

  RCSSYS=svn; export RCSSYS

  Define a name for your Dell EqualLogic device group in the LIST_OF_GROUPS configuration variable. In this example we'll use the name dell-storage:

  LIST_OF_GROUPS="dell-storage"; export LIST_OF_GROUPS

• Create the cloginrc configuration file, which will contain the login information for your Dell EqualLogic PS-Series devices and some default values:

  root@host:~$ touch /etc/rancid/cloginrc
  root@host:~$ chmod 660 /etc/rancid/cloginrc
  root@host:~$ chown root:rancid /etc/rancid/cloginrc
  root@host:~$ vi /etc/rancid/cloginrc

  Example:

  add user        dell-eql-1      dell-user
  add password    dell-eql-1      <login-passwort>
  
  [...]
  
  add user        *               <default-user>
  add password    *               <default-login-passwort>
  add method      *               ssh

  For the device named dell-eql-1 login as user dell-user with the password <login-passwort>.

  For all other systems, login as user <default-user> with the password <default-login-passwort>. The login method for all systems is via SSH.

  Since the cloginrc configuration file is parsed in a first-match fashion, the default values must always be at the bottom of the file.

• Add a new device type for Dell EqualLogic PS-Series storage arrays to the RANCID configuration. See man router.db and /etc/rancid/rancid.types.conf. In this example and in the general case of Dell Dell EqualLogic PS-Series storage arrays the name of the device type is equallogic:

  root@host:~$ vi /etc/rancid/rancid.types.conf

  Here we set the login script to be used to the new eqllogin. The postprocessing script is set rancid -t equallogic in order to call the new Perl module equallogic, which will do the actual processing. The command to be issued on the Dell EqualLogic device is set to save-config -verbose. The -verbose part is essential here, otherwise the configuration of the device will only be saved to a file on the Dell EqualLogic device and not be printed to the terminal:

  equallogic;login;eqllogin
  equallogic;script;rancid -t equallogic
  equallogic;module;equallogic
  equallogic;inloop;equallogic::inloop
  equallogic;command;equallogic::SaveConfiguration;save-config -verbose

• Change to the user rancid:

  root@host:~$ su - rancid

  • Create a symbolic link to the login configuration previously created in /etc/rancid/:

    rancid@host:~$ ln -s /etc/rancid/cloginrc /var/lib/rancid/.cloginrc

  • Initialize the directory structure for the RCS (CVS, SVN or Git) selected above. This will automatically be done for each device group configured in the LIST_OF_GROUPS configuration variable. The example shown here only creates the directory structure for the device group dell-storage defined above:

    rancid@host:~$ /usr/lib/rancid/bin/rancid-cvs
    Committed revision 1.
    Checked out revision 1.
    Updating '.':
    At revision 1.
    A         configs
    Adding         configs
    
    Committed revision 2.
    A         router.db
    Adding         router.db
    Transmitting file data .
    Committed revision 3.
    
    rancid@host:~$ find /var/lib/rancid/dell-storage/
    /var/lib/rancid/dell-storage
    /var/lib/rancid/dell-storage/configs
    /var/lib/rancid/dell-storage/router.db
    /var/lib/rancid/dell-storage/routers.all
    /var/lib/rancid/dell-storage/routers.down
    /var/lib/rancid/dell-storage/routers.up
    /var/lib/rancid/dell-storage/.svn
    /var/lib/rancid/dell-storage/.svn/entries
    /var/lib/rancid/dell-storage/.svn/format
    /var/lib/rancid/dell-storage/.svn/pristine
    /var/lib/rancid/dell-storage/.svn/pristine/da
    /var/lib/rancid/dell-storage/.svn/pristine/da/da39a3ee5e6b4b0d3255bfef95601890afd80709.svn-base
    /var/lib/rancid/dell-storage/.svn/tmp
    /var/lib/rancid/dell-storage/.svn/wc.db

  • Add Dell EqualLogic storage devices by their hostname to the configuration file router.db of the corresponding device group:

    rancid@host:~$ vi /var/lib/rancid/dell-storage/router.db

    In this example the device group dell-storage, the device type equallogic and the system dell-eql-1:

    dell-eql-1;equallogic;up;A comment describing the system dell-eql-1

  • Perform a login test with the previously configured new login script eqllogin for Dell EqualLogic devices on the newly defined system dell-eql-1. The following example output shows the steps that should automatically be performed by the eqllogin expect script. No manual intervention should be necessary.

    rancid@host:~$ /usr/lib/rancid/bin/eqllogin dell-eql-1
    spawn ssh -x -l grpadmin dell-eql-1
    The authenticity of host 'dell-eql-1 (<ip address>)' can't be established.
    RSA key fingerprint is <rsa key fingerprint>
    Are you sure you want to continue connecting (yes/no)?  yes
    Host dell-eql-1 added to the list of known hosts.
    Warning: Permanently added 'dell-eql-1,<ip address>' (RSA) to the list of known hosts.
    grpadmin@dell-eql-1's password: 
    Last login: Thu Dec  8 22:32:39 2016 from <ip address> on ttyp1
     
    
               Welcome to Group Manager
    
            Copyright 2001-2015 Dell Inc.
    
    
    
    dell-eql-1-grp> 

  • Finish the login test by manually logging out of the system:

    dell-eql-1-grp> logout
    Do you really want to logout? (y/n) [n]y
    
    dell-eql-1-grp> Connection to dell-eql-1 closed.

  • Manually perform an initial RANCID run to make sure everything works as expected:

    rancid@host:~$ rancid-run

    If everything ran successfully, there should now be a file /var/lib/rancid/dell-storage/configs/dell-eql-1 containing the output of the command save-config -verbose for the system dell-eql-1.

• Create the email aliases necessary for the proper delivery of the emails generated by RANCID. Again in this example for the device group dell-storage:

  root@host:~$ vi /etc/aliases

  rancid-dell-storage:       <email>@<domain>
  rancid-admin-dell-storage: <email>@<domain>

  Recreate your aliases DB. In case postfix is used as an MTA:

  root@host:~$ postalias /etc/aliases

• Enable the RANCID cron jobs. Adjust the execution times and intervals according to your needs:

  root@host:~$ vi /etc/cron.d/rancid

Some final words: The contents of the directories /var/lib/rancid/<device group>/ and /var/lib/rancid/<device group>/configs/ are maintained in the RCS – CVS, SVN or Git – of your choice. You can operate on those directories with the usual commands of the selected RCS. There are also some really nice and intuitive web frontends to the RCS of choice. For me, the combination of SVN as RCS and WebSVN as a web frontend worked out very well.

In the previous article Debugging Segfaults in Open-iSCSIs iscsiuio on Intel Broadwell i mentioned using a backported version of Open-iSCSI on Debian 8 (“Jessie”). This new post describes the backport and the changes provided by it in greater detail. All the changes to the original Debian package from “unstable” (“Sid”) can be found in my Debian Open-iSCSI Packaging repository on GitHub.

Starting point was a clone of the Debian Open-iSCSI Packaging repository at Git commit df150d90. Mind though, that in the meantime between creating the backport and writing this, the Debian Open-iSCSI maintainers have been busy and a more recent version of the Debian Open-iSCSI package from “unstable” (“Sid”) is now available.

Within this particular version of the Debian Open-iSCSI package, i first enabled the build of Open-iSCSIs iscsiuio. On the one hand, this was done in order to ensure that the iscsiuio code would successfully build even at this old level of the Open-iSCSI code. On the other hand, this would be used as a differentiator for any issues surfacing later on, after the move to the more recent upstream Open-iSCSI sources, indicating the root cause of those would then solely be with the newer upstream version of Open-iSCSI. Some integration into the general system environment was also added at this point. In detail the changes were:

• Git commit 32c96e6c removes the Debian patch 05-disable-iscsiuio.patch which disables the build of iscsiuio.

• Git commit 984344a1 enables the build of iscsiuio, extends the cleanup build targets and adds iscsiuio to the dh_systemd build targets.

• Git commit 89d845a9 adds the results from the successful build – the iscsiuio binary, the iscsiuio manual page, a readme file and a logrotate configuration file – to the Debian package. It also adds the kernel modules bnx2i and cnic to the list of kernel modules to be loaded at installation time.

• Git commit 89195bbe adds the systemd service and socket unit files for iscsiuio. Those files have been taken from this discussion on the Open-iSCSI mailing list and have slightly been altered.

With the above changes a intermediary package was build for testing purposes. During the following tests sometimes all currently mounted filesystems – even those distinctly not based on iSCSI volumes – would suddenly be unmounted. For some filesystems this would succeed, for others, like e.g. the /var and the root filesystem, this would fail due to them being currently in use. The issue particularly occured while stopping the open-iscsi service either via its regular init script or via its systemd service. This is usually done at system shutdown or during uninstall of the Open-iSCSI package. Tracking down the root cause of this issue led to an unhandled case in the umountiscsi.sh script, which is called while stopping the open-iscsi service. Specifically, the following code section is responsible for the observed behaviour:

debian/extra/umountiscsi.sh

256    if [ $HAVE_LVM -eq 1 ] ; then
257        # Look for all LVM volume groups that have a backing store
258        # on any iSCSI device we found. Also, add $LVMGROUPS set in
259        # /etc/default/open-iscsi (for more complicated stacking
260        # configurations we don't automatically detect).
261        for _vg in $(cd /dev ; $PVS --noheadings -o vg_name $iscsi_disks $iscsi_partitions $iscsi_multipath_disks $iscsi_multipath_partitions 2>/dev/null) $LVMGROUPS ; do
262            add_to_set iscsi_lvm_vgs "$_vg"
263        done

The heuristic of the umountiscsi.sh script are trying to identify iSCSI based disk devices which are valid candidates for proper deactivation upon system shutdown. It turned out that in LVM based setups where there are currently no iSCSI based disk devices present, the variables $iscsi_disks, $iscsi_partitions, $iscsi_multipath_disks and $iscsi_multipath_partitions are left empty by the scripts logic. In line 261 in the above code snippet, this leads to a call to the pvs --noheadings -o vg_name command without any additional arguments limiting its output of volume groups. Hence, the returned output is instead a complete list of all volume groups currently present on the system. Based on this list, the associated logical volumes for each volume group are determined and added to the list of devices to be unmounted. Finally all devices in this list are actually unmounted.

Without making too invasive changes to the script logic of umountiscsi.sh a quick'n'dirty solution was to introduce a check before the call to pvs which would determine whether the variables $iscsi_disks, $iscsi_partitions, $iscsi_multipath_disks and $iscsi_multipath_partitions are all empty. If this is the case, the call to pvs is simply skipped. The following patch shows the necessary code changes which are also available in Git commit 5118af7f:

umountiscsi.sh.patch

diff --git a/debian/extra/umountiscsi.sh b/debian/extra/umountiscsi.sh
index 1206fa1..485069c 100755
--- a/debian/extra/umountiscsi.sh
+++ b/debian/extra/umountiscsi.sh
@@ -258,9 +258,11 @@ enumerate_iscsi_devices() {
                # on any iSCSI device we found. Also, add $LVMGROUPS set in
                # /etc/default/open-iscsi (for more complicated stacking
                # configurations we don't automatically detect).
-               for _vg in $(cd /dev ; $PVS --noheadings -o vg_name $iscsi_disks $iscsi_partitions $iscsi_multipath_disks $iscsi_multipath_partitions 2>/dev/null) $LVMGROUPS ; do
-                       add_to_set iscsi_lvm_vgs "$_vg"
-               done
+               if [ -n "$iscsi_disks" -o -n "$iscsi_partitions" -o -n "$iscsi_multipath_disks" -o -n "$iscsi_multipath_partitions" ]; then
+                   for _vg in $(cd /dev ; $PVS --noheadings -o vg_name $iscsi_disks $iscsi_partitions $iscsi_multipath_disks $iscsi_multipath_partitions 2>/dev/null) $LVMGROUPS ; do
+                           add_to_set iscsi_lvm_vgs "$_vg"
+                   done
+               fi
 
                # $iscsi_lvm_vgs is now unique list
                for _vg in $iscsi_lvm_vgs ; do

After this was fixed, the last step was to finally move to the more recent upstream Open-iSCSI sources. In detail the changes in this last step were:

• Git commit f5ab51ff moves the code to version 2.0.873+git1.1dfb88a4 which is based upon the upstream Git commit 1dfb88a4. This is the last commit before the externalization of the Open-iSNS library. Since i didn't want to also backport the Open-iSNS packages from Debian “unstable” (“Sid”), i decided to just skip the next two upstream commits 76832662 and c6d1117b and stick with the locally delivered Open-iSNS library.

• Git commit 8c1e6974 removes the local Debian patches 01_spelling-errors-and-manpage-hyphen-fixes.patch, 02_make-iscsistart-a-dynamic-binary.patch and 03_respect-build-flags.patch which have already been merged into the more recent upstream Open-iSCSI sources. The remaining local Debian patches were renamed and reordered to 01_fix_iscsi_path.patch, 02_var-lock_var-run_transition.patch and 03_makefile_reproducibility_issues.patch. A whole bunch of new patches named {04,05,06,07,08,09,10,11,12,13,14,15}_upstream_git_commit_<Git commit ID>.patch were added in order to bring the sources up to the – by then most recent – upstream Git commit 0fa43f29.

• Git commit d051dece removes some files from the Debian package, which were dynamically generated during the build of iscsiuio.

• Finally Git commit 0fabb948 deals with the issue described in Debugging Segfaults in Open-iSCSIs iscsiuio on Intel Broadwell.

With the steps and changes described above, a backported version of Open-iSCSI using its most recent sources was created as a package for Debian 8 (“Jessie”). This package also supports offloaded iSCSI connections via the Broadcom BCM577xx and BCM578xx iSOEs with the use of iscsiuio. The package has been in production use for over a month now and no major issues – neither with the newer upstream Open-iSCSI sources, nor with use of Broadcom BCM577xx and BCM578xx iSOEs through iscsiuio – have emerged so far.

Open-iSCSIs tool iscsiuio, which is used to configure and manage Broadcom BCM577xx and BCM578xx iSCSI offload engines (iSOE), currently crashes with a segmentation fault upon target login on Intel Broadwell based systems. Comparable system setups, based on the older Intel Haswell chips do not show this issue.

In the past i've been using QLogic 4000 and QLogic 8200 Series network adapters and iSCSI HBAs which provide a full iSCSI offload engine (iSOE) implementation in the adapters firmware. Unfortunately the QLogic 8200 Series network adapters are no longer available for Dell M-Series blade servers. The alternatives offered by Dell are the Intel X520 and Intel X540 series adapters, or the Broadcom BCM57810S series adapters. Instead of using the Intel X520/X540, which provide no iSOE at all, i decided to go with the Broadcom BCM57810S, which at least provide some kind of iSOE. According to the VMware terminology the Broadcom BCM57810S are dependent hardware iSCSI Initiators. Dependent in this context means, that the iSOE does not implement all the necessary features and thus cannot perform all the tasks (e.g. TCP/IP stack, configuration and session management, authentication, etc.) necessary for target handling by itself. Rather, some of these tasks are provided by a third party on which this kind of iSOE depends on. In case of the Broadcom BCM57810S this third party is the iscsiuio daemon, which has for some time been part of the Open-iSCSI project. Simply put, the iscsiuio daemon acts as an intermediary between the iscsid on the one side and the QLogic¹⁾ NetXtreme II driver (kernel module bnx2 or bnx2x) and the QLogic²⁾ CNIC driver (kernel module cnic) on the other side, facilitating the creation and overall management of offloaded iSCSI sessions. Very simplified, the flow of information is as follows:

iscsiadm ←→ iscsid ←→ iscsiuio ←→ bnx2/bnx2x ←→ cnic ←→ Broadcom BCM57810S adapter ←→ Network ←→ Target

In my environment the Broadcom BCM57810S adapters are installed and used on six hosts (host1 and host{5,6,7,8,9}). They all connect to the same Dell EqualLogic storage systems in the backend, using the same network dedicated to iSCSI traffic. All hosts are Dell M630 blade servers with exactly the same firmware, operating system (Debian 8) and software versions. I'm using a backported version of Open-iSCSI, which is based on Git commit 0fa43f29, but excluding the commits 76832662 and c6d1117b which just implement the externalization of the Open-iSNS code. The systems originate from the same install image, so their configuration is – to a large extent – exactly the same. The only difference between the hosts is that host1 has Intel E5 v3 (aka Haswell) CPUs, while host{5,6,7,8,9} have Intel E5 v4 (aka Broadwell) CPUs.

On host1 everything works fine, iscsiuio runs as expected and access to targets via the Broadcom BCM57810S iSOEs is working flawlessly. On host{5,6,7,8,9} on the other hand, i was getting segmentation faults like the one in the example shown below, while trying to log in to any target.

host5:~# gdb /sbin/iscsiuio
GNU gdb (Debian 7.7.1+dfsg-5) 7.7.1
Copyright (C) 2014 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>.
Find the GDB manual and other documentation resources online at:
<http://www.gnu.org/software/gdb/documentation/>.
For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from /sbin/iscsiuio...(no debugging symbols found)...done.


(gdb) # run -d 4 -f
Starting program: /sbin/iscsiuio -d 4 -f
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
INFO  [Wed Jul 27 10:01:45 2016]Initialize logger using log file: /var/log/iscsiuio.log
INFO  [Wed Jul 27 10:01:45 2016]Started iSCSI uio stack: Ver 0.7.8.2
INFO  [Wed Jul 27 10:01:45 2016]Build date: Fri Jul 22 15:40:04 CEST 2016
INFO  [Wed Jul 27 10:01:45 2016]Debug mode enabled
INFO  [Wed Jul 27 10:01:45 2016]Running on sysname: 'Linux', release: '3.16.0-4-amd64', version '#1 SMP Debian 3.16.7-ckt
25-2+deb8u3 (2016-07-02)' machine: 'x86_64'
DBG   [Wed Jul 27 10:01:45 2016]Loaded nic library 'bnx2' Version: '0.7.8.2' build on Fri Jul 22 15:40:04 CEST 2016'
DBG   [Wed Jul 27 10:01:45 2016]Added 'bnx2' nic library
DBG   [Wed Jul 27 10:01:45 2016]Loaded nic library 'bnx2x' Version: '0.7.8.2' build on Fri Jul 22 15:40:04 CEST 2016'
DBG   [Wed Jul 27 10:01:45 2016]Added 'bnx2x' nic library
[New Thread 0x7ffff760f700 (LWP 4942)]
INFO  [Wed Jul 27 10:01:45 2016]signal handling thread ready
INFO  [Wed Jul 27 10:01:45 2016]nic_utils Found host[11]: host11
INFO  [Wed Jul 27 10:01:45 2016]Done capturing /sys/class/iscsi_host/host11/netdev
INFO  [Wed Jul 27 10:01:45 2016]Done capturing /sys/class/iscsi_host/host11/netdev
INFO  [Wed Jul 27 10:01:45 2016]nic_utils looking for uio device for eth3
WARN  [Wed Jul 27 10:01:45 2016]Could not find assoicate uio device with eth3
ERR   [Wed Jul 27 10:01:45 2016]nic_utils Could not determine UIO name for eth3
INFO  [Wed Jul 27 10:01:45 2016]nic_utils Found host[12]: host12
INFO  [Wed Jul 27 10:01:45 2016]Done capturing /sys/class/iscsi_host/host12/netdev
INFO  [Wed Jul 27 10:01:45 2016]Done capturing /sys/class/iscsi_host/host12/netdev
INFO  [Wed Jul 27 10:01:45 2016]nic_utils looking for uio device for eth2
INFO  [Wed Jul 27 10:01:45 2016]nic_utils eth2 associated with uio0
INFO  [Wed Jul 27 10:01:45 2016]nic_utils NIC not found creating an instance for host_no: 12 eth2
DBG   [Wed Jul 27 10:01:45 2016]Could not increase process priority: Success
[New Thread 0x7ffff6e0e700 (LWP 4943)]
DBG   [Wed Jul 27 10:01:45 2016]iscsi_ipc Started iscsid listening thread
DBG   [Wed Jul 27 10:01:45 2016]iscsi_ipc Waiting for iscsid command
INFO  [Wed Jul 27 10:01:45 2016]NIC_NL Netlink to CNIC on pid 4938 is ready
DBG   [Wed Jul 27 10:01:57 2016]iscsi_ipc recv iscsid request: cmd: 1, payload_len: 11720
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc Received request for 'eth2' to set IP address: '10.0.1.62' VLAN: '0'
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc Using netmask: 0.0.0.0
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc  eth2, using existing NIC
INFO  [Wed Jul 27 10:01:57 2016]nic_utils looking for uio device for eth2
INFO  [Wed Jul 27 10:01:57 2016]nic_utils eth2 associated with uio0
INFO  [Wed Jul 27 10:01:57 2016]Done capturing /sys/class/uio/uio0/name
INFO  [Wed Jul 27 10:01:57 2016]nic_utils eth2: Verified uio name bnx2x_cnic with library bnx2x
INFO  [Wed Jul 27 10:01:57 2016]eth2: found NIC with library 'bnx2x'
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2 library set using transport_name bnx2i
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2: requesting configuration using static IP address
DBG   [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2 couldn't find interface with ip_type: 0x2 creating it
INFO  [Wed Jul 27 10:01:57 2016]nic eth2: Added nic interface for VLAN: 0, protocol: 2
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2: created network interface
[New Thread 0x7ffff660d700 (LWP 4947)]
WARN  [Wed Jul 27 10:01:57 2016]nic_utils eth2: device already disabled: flag: 0x1088 state: 0x1
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2: configuring using static IP IPv4 address :10.0.1.62
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc  netmask: 255.255.255.0
[New Thread 0x7ffff5e0c700 (LWP 4948)]
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc ISCSID_UIP_IPC_GET_IFACE: command: 1 name: bnx2i.d0:43:1e:51:98:53, netdev: eth2 ipaddr: 10.0.1.62 vlan: 0 transport_name:bnx2i
INFO  [Wed Jul 27 10:01:57 2016]nic_utils eth2: spinning up thread for nic
DBG   [Wed Jul 27 10:01:57 2016]iscsi_ipc Waiting for iscsid command
[New Thread 0x7ffff560b700 (LWP 4949)]
DBG   [Wed Jul 27 10:01:57 2016]nic eth2: Waiting to be enabled
INFO  [Wed Jul 27 10:01:57 2016]Created nic thread: eth2
INFO  [Wed Jul 27 10:01:57 2016]iscsi_ipc eth2: started NIC enable thread state: 0x1
DBG   [Wed Jul 27 10:01:57 2016]nic eth2: is now enabled
INFO  [Wed Jul 27 10:01:57 2016]bnx2x eth2: bnx2x driver using version 1.78.19
ERR   [Wed Jul 27 10:01:58 2016]bnx2x /dev/uio0: uio device has been brought up via pid: 4938 on fd: 7
INFO  [Wed Jul 27 10:01:58 2016]Done capturing /sys/class/uio/uio0/name
INFO  [Wed Jul 27 10:01:58 2016]bnx2x eth2: Verified is a cnic_uio device
DBG   [Wed Jul 27 10:01:58 2016]bnx2x eth2: using rx ring size: 15, rx buffer size: 1024
INFO  [Wed Jul 27 10:01:58 2016]Done capturing /sys/class/uio/uio0/event
DBG   [Wed Jul 27 10:01:58 2016]bnx2x Chip ID: 168e1000
INFO  [Wed Jul 27 10:01:58 2016]nic_id eth2: is found at 03:00.00
INFO  [Wed Jul 27 10:01:58 2016]bnx2x eth2: func 0x0, pfid 0x0, client_id 0x88, cid 0x1
DBG   [Wed Jul 27 10:01:58 2016]bnx2x eth2: mode = 0x100
INFO  [Wed Jul 27 10:01:58 2016]bnx2x eth2:  Using mac address: d0:43:1e:51:98:53
INFO  [Wed Jul 27 10:01:58 2016]eth2: bnx2x initialized
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: Initialized ip stack: VLAN: 0
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: mac: d0:43:1e:51:98:53
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: Using IP address: 10.0.1.62
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: Using netmask: 255.255.255.0
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: enabled vlan 0 protocol: 2
INFO  [Wed Jul 27 10:01:58 2016]nic eth2: entering main nic loop
DBG   [Wed Jul 27 10:01:58 2016]nic_utils eth2: device enabled
[Thread 0x7ffff5e0c700 (LWP 4948) exited]
DBG   [Wed Jul 27 10:01:59 2016]iscsi_ipc recv iscsid request: cmd: 1, payload_len: 11720
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc Received request for 'eth2' to set IP address: '10.0.1.62' VLAN: '0'
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc Using netmask: 0.0.0.0
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc  eth2, using existing NIC
INFO  [Wed Jul 27 10:01:59 2016]nic_utils looking for uio device for eth2
INFO  [Wed Jul 27 10:01:59 2016]nic_utils eth2 associated with uio0
INFO  [Wed Jul 27 10:01:59 2016]eth2: Have NIC library 'bnx2x'
INFO  [Wed Jul 27 10:01:59 2016]Done capturing /sys/class/uio/uio0/name
INFO  [Wed Jul 27 10:01:59 2016]nic_utils eth2: Verified uio name bnx2x_cnic with library bnx2x
INFO  [Wed Jul 27 10:01:59 2016]eth2: found NIC with library 'bnx2x'
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc eth2 library set using transport_name bnx2i
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc eth2: requesting configuration using static IP address
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc eth2: using existing network interface
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc eth2: IP configuration didn't change using 0x2
INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc eth2: NIC already enabled flags: 0x1084 state: 0x4

INFO  [Wed Jul 27 10:01:59 2016]iscsi_ipc ISCSID_UIP_IPC_GET_IFACE: command: 1 name: bnx2i.d0:43:1e:51:98:53, netdev: eth2 ipaddr: 10.0.1.62 vlan: 0 transport_name:bnx2i
DBG   [Wed Jul 27 10:01:59 2016]iscsi_ipc Waiting for iscsid command
INFO  [Wed Jul 27 10:02:00 2016]NIC_NL Received path_req for host 12
INFO  [Wed Jul 27 10:02:00 2016]Done capturing /sys/class/iscsi_host/host12/netdev
DBG   [Wed Jul 27 10:02:00 2016]NIC_NL Pulled nl event
INFO  [Wed Jul 27 10:02:00 2016]NIC_NL eth2: Processing 'path_req'
DBG   [Wed Jul 27 10:02:00 2016]NIC_NL eth2: PATH_REQ with iface_num -1 VLAN 32768
DBG   [Wed Jul 27 10:02:00 2016]CNIC eth2: Netlink message with VLAN ID: 0, path MTU: 9000 minor: 0 ip_addr_len: 4
DBG   [Wed Jul 27 10:02:00 2016]CNIC eth2: src=10.0.1.62
DBG   [Wed Jul 27 10:02:00 2016]CNIC eth2: dst=10.0.1.2
DBG   [Wed Jul 27 10:02:00 2016]CNIC eth2: nm=255.255.255.0
INFO  [Wed Jul 27 10:02:00 2016]CNIC eth2: Didn't find IPv4: '10.0.1.2' in ARP table
DBG   [Wed Jul 27 10:02:00 2016]CNIC eth2: Sent cnic arp request for IP: 10.0.1.2
INFO  [Wed Jul 27 10:02:00 2016]Found 10.0.1.2 at b0:83:fe:cc:57:bb
DBG   [Wed Jul 27 10:02:00 2016]CNIC neighbor reply sent back to kernel 10.0.1.62 at b0:83:fe:cc:57:bb with vlan 0
INFO  [Wed Jul 27 10:02:00 2016]NIC_NL eth2: 'path_req' operation finished

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff660d700 (LWP 4947)]
__lll_unlock_elision (lock=0x55555577fd40, private=0) at ../nptl/sysdeps/unix/sysv/linux/x86/elision-unlock.c:29
29      ../nptl/sysdeps/unix/sysv/linux/x86/elision-unlock.c: No such file or directory.


(gdb) # info threads
  Id   Target Id         Frame
  6    Thread 0x7ffff560b700 (LWP 4949) "iscsiuio" 0x00007ffff76f1ae3 in select () at ../sysdeps/unix/syscall-template.S:81
* 4    Thread 0x7ffff660d700 (LWP 4947) "iscsiuio" __lll_unlock_elision (lock=0x55555577fd40, private=0) at ../nptl/sysdeps/unix/sysv/linux/x86/elision-unlock.c:29
  3    Thread 0x7ffff6e0e700 (LWP 4943) "iscsiuio" 0x00007ffff79c9ccd in accept () at ../sysdeps/unix/syscall-template.S:81
  2    Thread 0x7ffff760f700 (LWP 4942) "iscsiuio" do_sigwait (set=<optimized out>, sig=0x7ffff760eeac) at ../nptl/sysdeps/unix/sysv/linux/../../../../../sysdeps/unix/sysv/linux/sigwait.c:63
  1    Thread 0x7ffff7fea700 (LWP 4938) "iscsiuio" 0x00007ffff79c9e9d in recvmsg () at ../sysdeps/unix/syscall-template.S:81


(gdb) # thread apply all bt

Thread 6 (Thread 0x7ffff560b700 (LWP 4949)):
#0  0x00007ffff76f1ae3 in select () at ../sysdeps/unix/syscall-template.S:81
#1  0x000055555555ac06 in ?? ()
#2  0x000055555555d39e in nic_loop ()
#3  0x00007ffff79c30a4 in start_thread (arg=0x7ffff560b700) at pthread_create.c:309
#4  0x00007ffff76f887d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

Thread 4 (Thread 0x7ffff660d700 (LWP 4947)):
#0  __lll_unlock_elision (lock=0x55555577fd40, private=0) at ../nptl/sysdeps/unix/sysv/linux/x86/elision-unlock.c:29
#1  0x00007ffff79c7007 in pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:94
#2  0x000055555555e803 in nl_process_handle_thread ()
#3  0x00007ffff79c30a4 in start_thread (arg=0x7ffff660d700) at pthread_create.c:309
#4  0x00007ffff76f887d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

Thread 3 (Thread 0x7ffff6e0e700 (LWP 4943)):
#0  0x00007ffff79c9ccd in accept () at ../sysdeps/unix/syscall-template.S:81
#1  0x00005555555641b0 in ?? ()
#2  0x00007ffff79c30a4 in start_thread (arg=0x7ffff6e0e700) at pthread_create.c:309
#3  0x00007ffff76f887d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

Thread 2 (Thread 0x7ffff760f700 (LWP 4942)):
#0  do_sigwait (set=<optimized out>, sig=0x7ffff760eeac) at ../nptl/sysdeps/unix/sysv/linux/../../../../../sysdeps/unix/sysv/linux/sigwait.c:63
#1  0x00007ffff79ca693 in __sigwait (set=0x7ffff760eeb0, sig=0x0) at ../nptl/sysdeps/unix/sysv/linux/../../../../../sysdeps/unix/sysv/linux/sigwait.c:97
#2  0x000055555555a49c in _start ()

Thread 1 (Thread 0x7ffff7fea700 (LWP 4938)):
#0  0x00007ffff79c9e9d in recvmsg () at ../sysdeps/unix/syscall-template.S:81
#1  0x000055555555e5e9 in ?? ()
#2  0x000055555555eea8 in nic_nl_open ()
#3  0x000055555555a1b8 in main ()

The Open-iSCSI command which lead to this segmentation fault was a simple login at a previously defined target node:

host5:~# iscsiadm -m node -T <target iqn> -I <interface> --login

Double-checking the configuration, the firmware and software versions as well as the general hardware setup didn't yield any usable indication as to where the root cause of this issue might be. Searching the web for __lll_unlock_elision in conjunction with the pthread_* function calls, led me to the following resources:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=800574
https://lwn.net/Articles/534758/

Those are pointing towards a CPU (Broadwell and Skylake) specific problem when not carefully using mutexes. The general opinion from there and also other related bug reports seems to be, that the source of such issues is almost always an improper use of mutex locking, which – up to now – has either been tolerated or just by chance not lead to a failure. More recent CPUs and software implementations (e.g. the GNU libc) appear to be less forgiving in this regard. Thus the advice is to change the application behaviour towards a proper use of mutex locking, in order to address such an issue.

The article Intel's Broadwell Xeon E5-2600 v4 chips: So what's in it for you, smartie-pants coders offers a rather nice write-up of the new features introduced in the Intel Broadwell CPUs.

Tracking this issue further down in the Open-iSCSI sources, i ended up in the function nl_process_handle_thread() in iscsiuio/src/unix/nic_nl.c and specifically in the following code section:

iscsiuio/src/unix/nic_nl.c

474 /* NIC specific nl processing thread */
475 void *nl_process_handle_thread(void *arg)
476 {
[...]
483         while (!event_loop_stop) {
484                 char *data = NULL;
485
486                 rc = pthread_cond_wait(&nic->nl_process_cond,
487                                        &nic->nl_process_mutex);
488                 if (rc != 0) {
489                         LOG_ERR("Fatal error in NL processing thread "
490                                 "during wait[%s]", strerror(rc));
491                         break;
492                 }
[...]
499                 pthread_mutex_unlock(&nic->nl_process_mutex);
[...]

Debugging this revealed that the call to pthread_cond_wait() from the above GDB backtrace output of thread number 4 is the one from line 486 in the above code snippet.

Looking at the pthread_cond_wait() manpage showed the following constraint for its proper use:

[…]
The pthread_cond_timedwait() and pthread_cond_wait() functions shall
block on a condition variable. They shall be called with mutex locked
by the calling thread or undefined behavior results.
[…]

Although not shown in the above GDB output, this would on occasion – and again, probably just by chance – work on the first pass of the loop. At the end of the loop, at line 499 in the above code snippet, the mutex is then unlocked. Thus the cited constraint from the pthread_cond_wait() manpage is no longer met on the subsequent passes of the loop. On Intel E5 v3 (aka Haswell) CPUs, this seemed to be tolerated and without any impact. But on Intel E5 v4 (aka Broadwell) – and probably other CPUs implementing HLE and RTM – this causes the observed segmentation fault.

In order to verify my analysis and test this theory, i added a call to pthread_mutex_lock() right before the call to pthread_cond_wait() in line 486. The resulting change is available in the Git commit 9f770f9e of my Open-iSCSI fork on Github and also shown in the following patch:

nic_nl.c.patch

diff --git a/iscsiuio/src/unix/nic_nl.c b/iscsiuio/src/unix/nic_nl.c
index 391003f..581ddb0 100644
--- a/iscsiuio/src/unix/nic_nl.c
+++ b/iscsiuio/src/unix/nic_nl.c
@@ -483,6 +483,7 @@ void *nl_process_handle_thread(void *arg)
        while (!event_loop_stop) {
                char *data = NULL;
 
+               pthread_mutex_lock(&nic->nl_process_mutex);
                rc = pthread_cond_wait(&nic->nl_process_cond,
                                       &nic->nl_process_mutex);
                if (rc != 0) {

Posting this on the Open-iSCSI mailing list lead to this discussion. The suggestion from there was an additional change to the error handling code of nl_process_handle_thread() starting at line 488 in the above above code snippet. This adds proper handling of the locked mutex in case the loop is left due to an error returned from the call to pthread_cond_wait(). The resulting additional change is available in the Git commit 4191ca6b of my Open-iSCSI fork on Github. The following patch shows the summarised code changes:

nic_nl.c.patch

diff --git a/iscsiuio/src/unix/nic_nl.c b/iscsiuio/src/unix/nic_nl.c
index 391003f..1a920c7 100644
--- a/iscsiuio/src/unix/nic_nl.c
+++ b/iscsiuio/src/unix/nic_nl.c
@@ -483,9 +483,11 @@ void *nl_process_handle_thread(void *arg)
        while (!event_loop_stop) {
                char *data = NULL;
 
+               pthread_mutex_lock(&nic->nl_process_mutex);
                rc = pthread_cond_wait(&nic->nl_process_cond,
                                       &nic->nl_process_mutex);
                if (rc != 0) {
+                       pthread_mutex_unlock(&nic->nl_process_mutex);
                        LOG_ERR("Fatal error in NL processing thread "
                                "during wait[%s]", strerror(rc));
                        break;

With those two small changes to the sources of Open-iSCSIs iscsiuio, the iSCSI connections via the Broadcom BCM57810S iSOE do now work flawlessly even on newer Intel E5 v4 (aka Broadwell) based systems. Hopefully the original authors of the iscsiuio code at Broadcom/QLogic will also take part in the discussion and provide their feedback on the proposed code changes too.

The DokuWiki Taratasy Template offers a rather nice design, which i wanted to use for my new Wiki. This Wiki will be based upon the currently stable DokuWiki 2015-08-10a (aka “Detritus”) release. The Taratasy templates site currently shows the compatibility to the DokuWiki “Detritus” release as “unknown”, see the following screenshot:

Installing the Taratasy templates in an otherwise working DokuWiki 2015-08-10a instance lead to the rather broken appearance shown in the following example screenshot:

The colors and fonts are way off, the navigation bar at the head and bottom of the page as well as the tools menu are broken. Thus, my immediate suspicion for the root cause of the issues was in the direction of the CSS definitions provided by the Taratasy template.

Digging through the webserver and PHP logs as well as double checking the webservers rewrite rules showed no noticable indication of a problem with the DokuWiki instance itself. Switching to another template like e.g. Arctic also showed that the DokuWiki 2015-08-10a instance itself worked flawlessly. So this was not a general issue with my setup of the current DokuWiki release. The other way around, using the Taratasy template with an older DokuWiki 2014-09-29d (aka “Hrun”) release, also worked as expected and without any issues. So the problem was narrowed down to some change introduced between those two DokuWiki releases.

With the suspicion about the CSS definitions mentioned above, i tried to make some changes in the Taratasy templates style.ini file. Those changes immediately effected the layout of the re-rendered page. On the other hand, making some changes in the Taratasy templates style.local.ini file had no effect on the subsequently re-rendered page.

Browsing the commit history of the DokuWiki GitHub repository for the 2015-08-10a release and searching for commits with a reference to style.local.ini, i stubled upon the commit 656e584. The associated changes finally removed a feature which had already been marked as deprecated in previous DokuWiki releases. This feature was to include user defined CSS formatting definitions through the file style.local.ini in a templates main directory and merge these definitions with the ones from the templates style.ini file into the final CSS definitions. For several DokuWiki releases the user defined CSS formatting definitions are now supposed to be placed in the file DOKU_CONF.“tpl/$tpl/style.ini” instead.

With the source of the issue now known, my quick'n'dirty solution was to merge the CSS definitions from the templates style.local.ini into the templates style.ini and afterwards drop the now redundant style.local.ini. This was just easier to accomplish given my particular DokuWiki directory setup. Another solution would have been to create a directory DOKU_CONF/tpl/taratasy/ and move the style.local.ini into it as DOKU_CONF/tpl/taratasy/style.ini.

The modified template sources can be found in my Taratasy repository on GitHub

In conclusion, the following image shows an example screenshot of the same DokuWiki 2015-08-10a instance as above after the patch to the Taratasy template has been applied:

Dell PowerConnect and Dell PowerConnect M-Series switches can – with regard to their most important aspects like CPU, fans, PSU and temperature – already be monitored with the standard Check_MK distribution. This article introduces an enhanced version and additional Check_MK service checks to monitor additional aspects of Dell PowerConnect switches. It is targeted mainly towards the Dell PowerConnect M-Series switches used in Dell PowerEdge M1000e blade chassis, but can probably be used on standalone Dell PowerConnect switches as well.

For the impatient and TL;DR here is the Check_MK package of the enhanced version of the Dell PowerConnect monitoring checks:

Enhanced version of the Dell PowerConnect monitoring checks (Compatible with Check_MK versions 1.2.6 and earlier)
Enhanced version of the Dell PowerConnect monitoring checks (Compatible with Check_MK versions 1.2.8 and later)

The sources are to be found in my Check_MK repository on GitHub

The Dell PowerConnect M-Series switches to be used in Dell PowerEdge M1000e blade chassis – and possibly some newer Dell PowerConnect standalone switches too – are based on Broadcom FASTPATH silicon. While this hardware base introduces a plethora of other issues to be covered in detail in a separate article, it also introduces the possibility of breaking backwards compatibility with older Dell PowerConnect models from a monitoring point of view. Therefore, the new checks to cover the Broadcom FASTPATH based hardware were moved to a entirely new namespace. The file names of the new checks now use the prefix dell_powerconnect_bcm_ in contrast to the already existing stock Check_MK checks with their prefix dell_powerconnect_. Another difference to the stock Check_MK checks is the use of the FASTPATH Enterprise MIBs, which are specific to devices based on Broadcom silicon. The only exemptions are the checks dell_powerconnect_bcm_global_status and the dell_powerconnect_bcm_dnsstats, both monitor items which are not covered by the FASTPATH Enterprise MIBs.

All checks have been verified to work with the firmware versions 5.1.8.x and 5.1.9.x. For the newly introduced check dell_powerconnect_bcm_global_status the firmware version 5.1.9.4 or later is needed in order to avoid spurious error messages in the switch event log. See the section Additional Checks below for a more detailed explanation.

The discontinued, modified and additional checks are described in greater detail in the following three respective sections:

The two service checks dell_powerconnect_fans and dell_powerconnect_psu provided by the standard Check_MK distribution have become redundant for the Dell PowerConnect M-Series switches. The items to be monitored by both are not present in those devices, since the Dell PowerEdge M1000e blade chassis provides both central cooling and power supply facilities. Accordingly, the cooling and power supply facilities should be monitored via the Dell Chassis Managment Controller.

The two service checks dell_powerconnect_cpu and dell_powerconnect_temp have been renamed to dell_powerconnect_bcm_cpu and dell_powerconnect_bcm_temp respectively. They both have been modified to use the new Dictionary based parameters and factory settings for the CPU and temperature warning and critical levels. A SNMP example output for all OIDs used has been added to both service checks for documentation purposes. Manual pages, PNP4Nagios templates, WATO and Perf-O-Meter plugins have also been added for both service checks. With the added WATO plugins it is now possible to configure the CPU and temperature warning and critical levels through the WATO WebUI. The configuration options for the CPU levels can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Operating System Resources
         -> Dell PowerConnect CPU usage
            -> Create rule in folder ...
               [x] The levels for the overall CPU usage on Dell PowerConnect switches

The configuration options for the temperature levels can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Temperature, Humidity, Electrical Parameters, etc.
         -> Dell PowerConnect temperature
            -> Create rule in folder ...
               [x] Temperature levels for Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_cpu service check from the WATO WebUI:

Three average CPU utilization values for the time sample intervals 5, 60 and 300 seconds are checked. The Perf-O-Meter is split accordingly into three sections in order to be able to display all three average CPU utilization values at once.

The following image shows a status output example for the dell_powerconnect_bcm_temp service check from the WATO WebUI:

This example shows the status and the current values of the temperature sensors in a switch stack with two switch members.

The following two images show examples of PNP4Nagios graphs for both service checks:

The following table shows a condensed overview of the additional Check_MK service checks and their available components.

The first two columns should be pretty self-explanatory.

The Alarm column shows which checks will generate alarms based on the particular parameters monitored. Checks without an entry in the Alarm column are designed purely for long-term trends via their respective PNP4Nagios templates. All checks with an entry in the Alarm column use the new Dictionary based parameters and factory settings for their respective warning and critical levels. Where reasonable, those warning and critical levels are configurable through the WATO WebUI via an appropriate WATO plugin. See the last column, titled WATO plugin for the checks this applies to.

Manual pages are provided for each service check for documentation purposes. A SNMP example output is provided as a comment within the check script for all the OIDs used in the service check.

For all checks with an entry in the PNP4Nagios template column, a PNP4Nagios templates is provided in order to properly display the performance data delivered by the service check. Perf-O-Meter plugins are provided where reasonable, in order to display selected performance metrics in the service check overview of a host.

The specifics of each additional Check_MK service check are described in greater detail in the following sections.

The Check_MK service check dell_powerconnect_bcm_arp_cache monitors the current total number of entries in the ARP cache on Dell PowerConnect switches. This number is compared to either the default or configured warning and critical threshold values, and an alarm is raised accordingly. With the added WATO plugin it is possible to configure the warning and critical levels through the WATO WebUI and thus override the default values (warning: 3072; critical: 3584). The configuration options for the ARP cache levels can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Operating System Resources
         -> Dell PowerConnect ARP cache
            -> Create rule in folder ...
               [x] The levels for the number of ARP cache entries on Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_arp_cache service check from the WATO WebUI:

This example shows the current number of entries in the ARP cache along with the warning and critical threshold values.

In addition to the already mentioned total number of entries in the ARP cache, several other metrics are also collected as performance data. These are the overall ARP cache size, the number of static ARP entries and the peak values for both the current and the static number of ARP entries. The following image shows an example of the PNP4Nagios graph for the service check:

The Check_MK service check dell_powerconnect_bcm_cos_queue monitors the number of packets dropped at each CoS queue for the CPU (quoted from the FASTPATH Enterprise MIB). Unfortunately the only other description available in the FASTPATH Enterprise MIBs is almost as cryptic as the first one: Number of packets dropped at this CPU CoS queue because the queue was full. The metric probably relates to the switches Class of Service (CoS) feature in a Quality of Service (QoS) setup. Currently, the dell_powerconnect_bcm_cos_queue service check is used purely for long-term trends via its respective PNP4Nagios template and thus only gathers its metrics as performance data.

The following image shows a status output example for the dell_powerconnect_bcm_cos_queue service check from the WATO WebUI:

The following image shows an example of the PNP4Nagios graph for the service check:

The Check_MK service check dell_powerconnect_bcm_cpu_proc monitors the same CPU utilization metrics as the previously described dell_powerconnect_bcm_cpu service check, but on a more detailed, per process level. The three average CPU utilization values for the time sample intervals 5, 60 and 300 seconds are for each process compared to either the default or configured warning and critical threshold values and an alarm is raised accordingly. There is currently the limitation in the checks logic that warning and critical threshold values apply globally to all processes. Individual warning and critical threshold values for each process are currently not supported. With the added WATO plugin it is possible to configure the warning and critical levels through the WATO WebUI and thus override the default values (warning: 80%; critical: 90%) for the average CPU utilization. The configuration options for the per process CPU utilization levels can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Operating System Resources
         -> Dell PowerConnect CPU usage (per process)
            -> Create rule in folder ...
               [x] The levels for the per process CPU usage on Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_cpu_proc service check from the WATO WebUI:

The following image shows only four examples of PNP4Nagios graphs for the service checks:

The selected example graphs show the average CPU utilization over the 5, 60 and 300 seconds time sample intervals for the processes SNMPTask, bcmRX, dot1s_timer_task and osapiTimer. Mind though that this is only a small selection of the various processes that can be found running on the Broadcom FASTPATH based Dell PowerConnect switches. Some processes are always to be found, others appear only after a specific feature – covered by appropriate processs – is enabled on the switch. Unfortunately i've not been able to find a complete list of the possible processes nor a good and comprehensive description of the purpose of each process. Sometimes – like in the case of the process SNMPTask – the purpose can be guessed from process name. So overall i'd say the per process CPU utilization metric is probably best used as a metric for long-term trends in conjunction with support from Broadcom or Dell, when dealing with a specific issue on the switch or an unusually high CPU utilization of a specific process.

The Check_MK service check dell_powerconnect_bcm_dnsstats monitors various aspects and metrics of the switches local DNS resolver. The metrics gathered can be grouped into three categories:

• DNS Resolver: The number of DNS resolver queries and the number of DNS responses to those queries. For the DNS responses the number of responses in each response category. The response categories are: Non-auth Answers, Non-auth No-answer, Received Responses, Unparsable Responses, Martians Responses and Fallbacks.

• DNS Resolver RCODE: The number of DNS resolver responses by resonse code. See 1035 for the details on DNS response codes.

• DNS Cache: The number of DNS resouce records that have been successfully added or have failed to be added to the DNS resolver cache.

See 1612, 1035 and the service checks man page for a detailed description of the metrics covered by the dell_powerconnect_bcm_dnsstats service check. Currently, the dell_powerconnect_bcm_dnsstats service check is used purely for long-term trends via its respective PNP4Nagios template and thus only gathers its metrics as performance data.

The following image shows a status output example for the dell_powerconnect_bcm_dnsstats service check from the WATO WebUI:

The following image shows an example of the three PNP4Nagios graphs for the service check:

The Check_MK service check dell_powerconnect_bcm_global_status monitors just one metric, the productStatusGlobalStatus from the Dell Vendor MIB for PowerConnect devices. As the name of the metric suggests, it represents an aggregated global status for a Dell PowerConnect device. The global status can assume one of the three values, shown in the following table:

While the information about the fan and PSU status is redundant for the Dell PowerConnect M-Series switches, the information about hard- and software error conditions might be quite valueable.

When we first implemented the enhanced version of the Dell PowerConnect monitoring checks, we noticed spurious error messages suddenly appearing in the switch event log and subsequently in our syslog servers. The messages showing up looked like the following example:

<189> OCT 14 12:48:50 <Management IP address>-1 MGMT_ACAL[251047504]: macal_api.c(873) 38462 %% macalRuleActionGet(): List  does not exist.

Disabling one check after another, we narrowed the source of this error message down to the dell_powerconnect_bcm_global_status service check. Logging a support case with Dell eventually lead to the following explaination from Dell PowerConnect engineering:

Hi Frank,

I got an update from our engineering team and they can see the problem
when snmpwalk is executed against switch but issue is not seen if snmpget
is executed on all OIDs.

They are working on a fix.

Once fix is available it will be included in the next FW patch release
for this switch. […]

At the time we were running the newest available firmware, which back then was version 5.1.9.3. After updating to the firmware version 5.1.9.4 which was released later on, the above error messages stopped showing up.

The Check_MK service check dell_powerconnect_bcm_ip_conflict monitors the status of the built-in IP address conflict detection feature of a Dell PowerConnect switch. If an IP address conflict is detected, an alarm with the status warning is raised. In addition to the alarm status, the service check will also report the conflicting IP, (if available) the MAC address of the device causing the conflict and the date and time the conflict was detected. The last bit of information is relative to the switches date and time settings. Needless to say, a properly configured date and time or a time synchronisation via NTP on the siwtch is quite helpful in such a case.

Once an IP address conflict is detected by a Dell PowerConnect switch, this status will not resolve itself automatically or time out in any way. The issue has to be acknowledged manually on the Dell PowerConnect switch. This can be achieved e.g. on the switchs' CLI with the following commands:

switch> enable
switch# clear ip address-conflict-detect

The Check_MK service check dell_powerconnect_bcm_logstats monitors several metrics of the logging facility on Dell PowerConnect switches. These are the:

• total number of log messages received by the log process, including dropped and ignored messages.

• number of dropped log messages, which could not be processed by the log process due to an error or lack of resources.

• number of relayed log messages. These are log messages which have been forwarded to a remote syslog host by the log process. If multiple remote syslog hosts are configured, each message is counted multiple times, once for each of the configured syslog hosts.

Currently, the dell_powerconnect_bcm_logstats service check is used purely for long-term trends via its respective PNP4Nagios template and thus only gathers its metrics as performance data.

The following image shows a status output example for the dell_powerconnect_bcm_logstats service check from the WATO WebUI:

The following image shows an example of the PNP4Nagios graph for the service check:

Unfortunately the information as to why log messages might have been erroneous or which resources (CPU cycles, free memory, etc.) were missing at the time of processsing the log message is scarce. The metrics about the logging facility are therefore – again – probably best used as metrics for long-term trends in conjunction with support from Broadcom or Dell, when dealing with a specific issue on the switch.

The Check_MK service check dell_powerconnect_bcm_mbuf monitors two groups of metrics regarding the memory or message buffers on Dell PowerConnect switches. The first group is the overall number of currently available memory or message buffers on the switch. This group consists of just one metric. The second group is the number of total and the number of failed memory or message buffer allocation attempts for packets arriving at the switches CPU. Those two metrics are gathered for each of memory or message buffer classes. The names of the currently available memory or message buffer classes are “Transmit”, “Rx High”, “Rx Mid0”, “Rx Mid1”, “Rx Mid2” and “Rx Normal”.

The dell_powerconnect_bcm_mbuf service check is currently used only for long-term trends via its respective PNP4Nagios template and thus only gathers its metrics as performance data.

The following image shows a status output example for the dell_powerconnect_bcm_mbuf service check from the WATO WebUI:

The following image shows an example of the seven PNP4Nagios graphs for the service check. One graph for the overall available memory or message buffers and one graph for the allocation attempts on each of the six memory or message buffer classes:

Similarly to the process names described in the previous section Process CPU Usage, i've also not been able to find a good and comprehensive description of the memory or message buffer classes defined on the Broadcom FASTPATH based Dell PowerConnect switches. Some meaning can again be derived from the name of the particular memory or message buffer class, but it is much more limited than in case of the process names. Beyond that, questions like the following – but not limited to – immediately come to mind:

• which type of packets are forwarded to the CPU instead of being directly processed by the switching silicon of the device?

• why are there several receive classes (“Rx …”) but only one transmit class?

• what is the difference between the multiple receive classes and by what algorithm are packets assigned to a specific receive class?

• what is likely the root cause of a failed memory or message buffer allocation attempt?

• what are the effects of a failed memory or message buffer allocation attempt. Are packets going to be dropped due to this, or is the allocation attempt retried?

• what design, implementation and configuration options should be taken into consideration in order to avoid failed memory or message buffer allocation attempts?

Unfortunately they remain unanswered due to the lack of comprehensive documentation. The metrics regarding the memory or message buffers are therefore – again – probably best used for long-term trends in conjunction with support from Broadcom or Dell, when dealing with a specific issue on the switch.

The Check_MK service check dell_powerconnect_bcm_memory monitors the current memory (RAM) usage on Dell PowerConnect switches. The amount of currently free memory is compared to either the default or configured warning and critical threshold values, and an alarm is raised accordingly. With the added WATO plugin it is possible to configure the warning and critical levels through the WATO WebUI and thus override the default values (warning: 51200 KBytes; critical: 25600 KBytes of free memory). The configuration options for the free memory levels can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Operating System Resources
         -> Dell PowerConnect memory usage
            -> Create rule in folder ...
               [x] The levels for the amount of free memory on Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_memory service check from the WATO WebUI:

This example shows the current amount of free memory and the total memory size both measured in kilobytes.

The following image shows an example of the PNP4Nagios graph for the service check:

The Check_MK service check dell_powerconnect_bcm_sntp monitors the current status of the SNTP client on Dell PowerConnect switches. In order to achieve this, the check iterates over the list of SNTP servers configured as time references for the SNTP client on the switch. For each configured SNTP server, the status of the last connection attempt from the SNTP client on the switch to that particular SNTP server is evaluated. The overall number of SNTP servers with a connection status equal to success is counted and this number is compared to either the default or configured warning and critical threshold values, and an alarm is raised accordingly. With the added WATO plugin it is possible to configure the warning and critical levels through the WATO WebUI and thus override the default values (warning: 1 ; critical: 0 servers successfully connected). The configuration options for the levels of successful SNTP server connections can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Applications, Processes & Services
         -> Dell PowerConnect SNTP status
            -> Create rule in folder ...
               [x] Successful SNTP server connections on Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_sntp service check from the WATO WebUI:

This example shows the current status of the SNTP client which has successfully connected the one configured SNTP server.

In addition to the aggregated current connection status of the SNTP client to all configured SNTP servers, two other metrics are – for each SNTP server – collected as performance data. These are the overall number of SNTP requests – including retries – and the number of failed SNTP requests the client made to a particular SNTP server. The following image shows an example of the PNP4Nagios graph for the service check:

The Check_MK service check dell_powerconnect_bcm_ssh_sessions monitors just one metric, the number of currently active SSH sessions on the Dell PowerConnect device. This number is compared to either the default or configured warning and critical threshold values, and an alarm is raised accordingly. With the added WATO plugin it is possible to configure the warning and critical levels through the WATO WebUI and thus override the default values (warning: 5; critical: 5 active SSH sessions). The configuration options for the number of active SSH sessions can be found under:

-> Host & Service Parameters
   -> Parameters for discovered services
      -> Applications, Processes & Services
         -> Dell PowerConnect SSH sessions
            -> Create rule in folder ...
               [x] Active SSH sessions on Dell PowerConnect switches

The following image shows a status output example for the dell_powerconnect_bcm_ssh_sessions service check from the WATO WebUI:

This example shows the current number of active SSH sessions along with the warning and critical threshold values.

The following image shows an example of the PNP4Nagios graph for the service check:

Adding the enhanced version of the Dell PowerConnect monitoring checks to your Check_MK server enables you to monitor various additional aspects of your Dell PowerConnect devices. New Dell PowerConnect devices should pick up the additional service checks immediately. Existing Dell PowerConnect devices might need a Check_MK inventory to be run explicitly on them in order to pick up the additional service checks.

Along with the built-in Check_MK monitoring of interfaces of network equipment, the monitoring of services (SSH, HTTP and HTTPS) and the status of certificates, as well as the previously described monitoring of RMON Interface Statistics, this enhanced version of the Dell PowerConnect monitoring checks enables you to create a complete monitoring solution for your Dell PowerConnect M-Series switches.

I hope you find the provided new and enhanced checks useful and enjoyed reading this blog post. Please don't hesitate to drop me a note if you have any suggestions or run into any issues with the provided checks.