Puppet users - May 2008 - connection timeout / memory usage / locks / recompiling

Hi

I have created a module for djbdns [1], which will generate the data
file (dns entries) from defines in puppet classes. This is all fine and
working. :)

However I autogenerated after a few tests the classes for the domains
from the current source of our domains. Which are about 50 classes each
calling around 3 defines, which will be concatenated_file calls (from
davids common module), which are in the end around 350 files (done by
the file-resource and using the content parameter), often containing
only a single line.

After having all these classes I tried to run puppet once with puppetd
--test on the client which then should get all these classes down to client.
After running it the first time, I got a timeout, I remembered that
there was a fix in the upcoming release, which will respect the timeouts
set by config. Applied the patch from the ticket #1176 [2] on the client
and rerun the client with puppetd --test. No it was running quite longer :)

BUT it is still failling:

time puppetd --test
info: Loading fact acpi_available
info: Loading fact interfaces
info: Loading fact netmask
info: Loading fact virtual
info: Loading fact configured_ntp_servers
info: Loading fact selinux
info: Retrieving plugins
notice: Ignoring cache
err: Could not call puppetmaster.getconfig: #<RuntimeError: HTTP-Error:
504 Gateway Time-out>
err: Could not retrieve catalog: HTTP-Error: 504 Gateway Time-out
warning: Not using cache on failed catalog

real	4m23.383s
user	1m10.460s
sys	3m8.400s

I''m using the setup with nginx described on the wiki [3]

I then tried to adjust ssl_session_timout and ssl_session_cache with
giving them 10m more. However this didn''t change anything. One run was
aborted after the same amount of time.

I realized then on the master that it used a hell lot more of memory and
that it was nearly by trashing. So I examined more the master and saw,
that one of the 4 puppetmaster-daemons (the one which the problemhost
connected to) had a fast growing memory usage during the problemhost was
connected, at the end it had a bit more than 30% memory usage (xen
instance has about 1.4G RAM, 3G swap, and 2.33GHz cpu, which isn''t
really used by something else). After the client disconnected with the
above error, I could see that the process was still running high on cpu
and even consuming more memory, but it grew then only about 3%.

I also the could see strange behaviour in the logs:

Sun May 18 10:05:30 +0200 2008 Puppet (notice): Compiled catalog for
host1 in 18.92 seconds
Sun May 18 10:08:52 +0200 2008 Puppet (notice): Compiled catalog for
puppetmaster in 5.92 seconds
Sun May 18 10:10:01 +0200 2008 Puppet (notice): Compiled catalog for
host2 in 4.06 seconds
Sun May 18 10:14:59 +0200 2008 Puppet (err): Could not store configs:
Mysql::Error: Lock wait timeout exceeded; try restarting transac
tion:             DELETE FROM `param_values`
            WHERE `id` = 4880339
Sun May 18 10:14:59 +0200 2008 Puppet (notice): Compiled catalog for
problemhost in 157.61 seconds
Sun May 18 10:17:35 +0200 2008 Puppet (err): Could not store configs:
Mysql::Error: Lock wait timeout exceeded; try restarting transac
tion:             DELETE FROM `param_values`
            WHERE `id` = 4880339
Sun May 18 10:17:37 +0200 2008 Puppet (notice): Compiled catalog for
problemhost in 255.58 seconds
Sun May 18 10:27:42 +0200 2008 Puppet (err): Could not store configs:
Mysql::Error: Lock wait timeout exceeded; try restarting transac
tion:             DELETE FROM `param_values`
            WHERE `id` = 4880339
Sun May 18 10:27:42 +0200 2008 Puppet (notice): Compiled catalog for
problemhost in 800.52 seconds
Sun May 18 10:41:32 +0200 2008 Puppet (notice): Compiled catalog for
problemhost in 1809.78 seconds
Sun May 18 10:44:20 +0200 2008 Puppet (notice): Compiled catalog for
host2 in 108.01 seconds
Sun May 18 10:47:06 +0200 2008 Puppet (err): no classes for this kernel
yet defined! at /srv/puppet/development/modules/ntp/manifests/
init.pp:21 on node host2
Sun May 18 10:47:06 +0200 2008 Puppet (err): no classes for this kernel
yet defined! at /srv/puppet/development/modules/ntp/manifests/
init.pp:21 on node host2
Sun May 18 10:47:17 +0200 2008 Puppet (err): no classes for this kernel
yet defined! at /srv/puppet/development/modules/ntp/manifests/
init.pp:21 on node host2
Sun May 18 10:47:17 +0200 2008 Puppet (err): no classes for this kernel
yet defined! at /srv/puppet/development/modules/ntp/manifests/
init.pp:21 on node host2
Sun May 18 10:48:05 +0200 2008 Puppet (notice): Compiled catalog for
host2 in 201.16 seconds
Sun May 18 10:48:10 +0200 2008 Puppet (notice): Compiled catalog for
puppetmaster in 206.64 seconds
Sun May 18 10:54:44 +0200 2008 Puppet (notice): Compiled catalog for
puppetmaster in 722.00 seconds
Sun May 18 10:55:06 +0200 2008 Puppet (notice): Compiled catalog for
puppetmaster in 691.28 seconds
Sun May 18 11:17:32 +0200 2008 Puppet (notice): Caught TERM; shutting down

So if I start the puppetd --test (after a previous restart of the
master) only _once_, I can see the following behaviour:

1. puppetmaster''s memory is growing very fast and very quick, up to a
not yet known level.
2. client (problemhost in the logs) exits with a gateway timeout, which
can''t be adjusted by reconfiguring nginx. (or did I miss anything?
maybe
a firewall issue?)
3. puppetmaster seems still to be compiling the client''s manifests.
4. puppetmaster is getting some locking issues with the mysql db
5. puppetmaster is _recompiling_ the client''s manifests several times
without having the client reconnecting again.
6. compiling of other hosts take afterwards 10 times as before and they
start complaining about things which are simply not true: the ntp error
is a defined fail if the found kernel by facter is not linux or openbsd
(i didn''t configure ntp yet on any other system) and this worked before
as you can see in the logs! I assume this is realted to some corrupted
memory?
7. client''s manifest compiling will still take ages until i restart the
master.

So in my opinion there seems to be still some issues with memory or
whatever. Actually I don''t see any misconfiguration on my site, nor are
the generation of that much files in my opinion really a problem? Or are
they and this is the same as with serving huge and a lot of files (I
thought that maybe serving and using the content param are different)
and people should wait until the REST support is here and then ideas
like these will work? ;)

Or are there any other ideas what could be wrong and how I could fix
this issue?

Ah and I use 0.24.4 on all clients, as well the master. The master has
the latest nagios_commands improvements applied (for some other host)
and the problemclient has the timeout patch applied.

thanks a lot and greets pete.

[1] http://github.com/duritong/puppet-djbdns/tree/master
[2] http://reductivelabs.com/trac/puppet/ticket/1176
[3] http://reductivelabs.com/trac/puppet/wiki/UsingMongrelNginx

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

On May 18, 2008, at 6:29 AM, Peter Meier wrote:
> So in my opinion there seems to be still some issues with memory or
> whatever. Actually I don''t see any misconfiguration on my site,
nor
> are
> the generation of that much files in my opinion really a problem? Or  
> are
> they and this is the same as with serving huge and a lot of files (I
> thought that maybe serving and using the content param are different)
> and people should wait until the REST support is here and then ideas
> like these will work? ;)
I can''t disagree with your conclusion -- there''s a problem
there.

The next step is to figure out what''s causing the problem and fix it.

-- 
Most people are born and years later die without really having lived
at all. They play it safe and tiptoe through life with no aspiration
other than to arrive at death safely. -- Tony Campolo, "Carpe Diem"
---------------------------------------------------------------------
Luke Kanies | http://reductivelabs.com | http://madstop.com


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

Hi
> I can''t disagree with your conclusion -- there''s a
problem there.
> 
> The next step is to figure out what''s causing the problem and fix
it.
could somebody point me what these steps would be? I''d like to do it
and
nail down the problem. I would have time now to investigate on that problem.

However another idea is that the ruby on centos5 is quite old and might
be related to the problems. And remembering that @ work we had some
weird similar problems with an own ruby script, which disappeared after
upgrading ruby to 1.9 this might be also worth a try?

greets Pete

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

On May 22, 2008, at 5:26 AM, Peter Meier wrote:
>
> Hi
>
>> I can''t disagree with your conclusion -- there''s a
problem there.
>>
>> The next step is to figure out what''s causing the problem and
fix it.
>
> could somebody point me what these steps would be? I''d like to do
it
> and
> nail down the problem. I would have time now to investigate on that  
> problem.
>
> However another idea is that the ruby on centos5 is quite old and  
> might
> be related to the problems. And remembering that @ work we had some
> weird similar problems with an own ruby script, which disappeared  
> after
> upgrading ruby to 1.9 this might be also worth a try?

An old ruby could cause problems, definitely, but... I haven''t heard  
of anyone running Puppet with 1.9 (is it even officially out yet?).

Catch me next week, maybe on IRC, and I''ll be in a better position to  
help you track it down, hopefully.

-- 
All power corrupts, but we need the electricity.
    -- Unknown
---------------------------------------------------------------------
Luke Kanies | http://reductivelabs.com | http://madstop.com


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

Luke Kanies wrote:
> An old ruby could cause problems, definitely, but... I haven''t
heard
> of anyone running Puppet with 1.9 (is it even officially out yet?).
>
>   
Sorta related... Is anyone running puppet on ruby v1.6 or will this work?
That is what comes stock with Mac OS X 10.3.


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

On Thu, May 22, 2008 at 8:10 AM, Mark Foster <mfoster@bitpusher.com>
wrote:
> Sorta related... Is anyone running puppet on ruby v1.6 or will this work?
> That is what comes stock with Mac OS X 10.3.
Ruby 1.6 is fine.

Adam

-- 
HJK Solutions - We Launch Startups - http://www.hjksolutions.com
Adam Jacob, Senior Partner
T: (206) 508-4759 E: adam@hjksolutions.com

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

Hi
> An old ruby could cause problems, definitely, but... I haven''t
heard
> of anyone running Puppet with 1.9 (is it even officially out yet?).
nope, we took the vanilla one.
> Catch me next week, maybe on IRC, and I''ll be in a better position
to
> help you track it down, hopefully.
ok I''ll do so around thursday.

thanks and greets pete


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to puppet-users@googlegroups.com
To unsubscribe from this group, send email to
puppet-users-unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en
-~----------~----~----~----~------~----~------~--~---