Xymon Mailing List Archive search

memory status not updated on some Solaris clients after upgrade

3 messages in this thread

list Andy Smith · Wed, 15 Jan 2014 10:16:00 +0000 ·
Hi,

Since upgrading to 4.3.13, the 'memory' tests on some Solaris clients have
turned purple.  All the other tests for these hosts are updated correctly,
but svcstatus for memory reflects that no update has been applied and the
rrd files memory.real and memory.swap are not being updated.  None of the
clients were upgraded, only the server.

The clientlog for the sections memory or swap for a solaris host that works
and a solaris host that doesnt work are indistinguishable in format.

Whilst checking the timestamp on the memory.real.rrd file, I noticed that
some of the rrd files in both directories are up to 60 minutes old, but if
an attempt to access the data is made by for instance opening the status
page, all the rrd files seem to get updated, so I am not sure this is
relevant, but if anyone can explain it, I would be grateful.

Does anyone have any suggestions how I can debug the errant solaris memory
tests?  There is nothing relevant I can find in any of the logs which is
tagged with the host name or even the word 'memory'.  I have upgraded one
solaris client (4.3.10->4.3.13) just to rule it out, but as expected, it
made no difference.

Thanks
-- 
Andy
list Andy Smith · Fri, 17 Jan 2014 20:20:55 +0000 ·
Hi,
quoted from Andy Smith

On 15 January 2014 10:16, Andy Smith <user-982f5f6d4d28@xymon.invalid> wrote:
Hi,

Since upgrading to 4.3.13, the 'memory' tests on some Solaris clients have
turned purple.  All the other tests for these hosts are updated correctly,
but svcstatus for memory reflects that no update has been applied and the
rrd files memory.real and memory.swap are not being updated.  None of the
clients were upgraded, only the server.

The clientlog for the sections memory or swap for a solaris host that
works and a solaris host that doesnt work are indistinguishable in format.

Whilst checking the timestamp on the memory.real.rrd file, I noticed that
some of the rrd files in both directories are up to 60 minutes old, but if
an attempt to access the data is made by for instance opening the status
page, all the rrd files seem to get updated, so I am not sure this is
relevant, but if anyone can explain it, I would be grateful.

Does anyone have any suggestions how I can debug the errant solaris memory
tests?  There is nothing relevant I can find in any of the logs which is
tagged with the host name or even the word 'memory'.  I have upgraded one
solaris client (4.3.10->4.3.13) just to rule it out, but as expected, it
made no difference.

Thanks
--
Andy
If a Solaris host is stood up with no swap configured, any memory checks
submitted to 4.3.13 (and possibly previous versions) are silently
discarded.  This affects most (but not all) of our non-global zones.  This
patch for xymond/client/solaris.c fixes the problem.

On the subject of non-global zones, it is necessary to run slightly
different commands in a non-global zone to extract the correct
information.  This is especially true in respect of memory and swap and
will be particularly visible to anyone running memory capped zones.  I
offer this patch to xymonclient-sunos.sh if anyone is interested.
Thanks
-- 
Andy
Attachments (2)
list Henrik Størner · Sun, 19 Jan 2014 22:28:11 +0100 ·
quoted from Andy Smith
Den 17-01-2014 21:20, Andy Smith skrev:
If a Solaris host is stood up with no swap configured, any memory checks
submitted to 4.3.13 (and possibly previous versions) are silently
discarded.  This affects most (but not all) of our non-global zones.
This patch for xymond/client/solaris.c fixes the problem.

On the subject of non-global zones, it is necessary to run slightly
different commands in a non-global zone to extract the correct
information.  This is especially true in respect of memory and swap and
will be particularly visible to anyone running memory capped zones.  I
offer this patch to xymonclient-sunos.sh if anyone is interested.
Thanks, applied. I don't know enough about Solaris to determine if your 
client-script patches may break some setups, so I trust that you have 
tested it on different systems.

The change for server-side handling of the swap data seems sound, so I 
have no worries about that.


Regards,
Henrik