I'm having a similar problem.
I know for a fact that I am not contending with other agents, because this
is in a lab environment and none of my agents is scheduled for periodic
runs (yet).
I have successfully run puppet agent -t first time, signed the cert, and
run it a second time to pull over stdlib and other modules on agents
running RHEL 6 and Solaris 10u10 x86.
But I'm getting this timeout on a Solaris 10u10 box running on a T4-1 SPARC
system.
This was my third run:
# date;puppet agent -t;date
Wed Sep 4 14:12:05 CDT 2013
Info: Retrieving plugin
Notice: /File[/var/lib/puppet/lib/puppet/parser/functions/count.rb]/ensure:
defined content as '{md5}9eb74eccd93e2b3c87fd5ea14e329eba'
Notice:
/File[/var/lib/puppet/lib/puppet/parser/functions/validate_bool.rb]/ensure:
defined content as '{md5}4ddffdf5954b15863d18f392950b88f4'
Notice:
/File[/var/lib/puppet/lib/puppet/parser/functions/get_module_path.rb]/ensure:
defined content as '{md5}d4bf50da25c0b98d26b75354fa1bcc45'
Notice:
/File[/var/lib/puppet/lib/puppet/parser/functions/is_ip_address.rb]/ensure:
defined content as '{md5}a714a736c1560e8739aaacd9030cca00'
Error:
/File[/var/lib/puppet/lib/puppet/parser/functions/is_numeric.rb]/ensure:
change from absent to file failed: execution expired
Error: Could not retrieve plugin: execution expired
Info: Caching catalog for AGENT
Info: Applying configuration version '1378322110'
Notice: Finished catalog run in 0.11 seconds
Wed Sep 4 14:15:58 CDT 2013
Each time I've run it, I get about 10 or so files and then I get "execution
expired".
What I'd really like to see is whether I can increase the expiry timeout.
Some other details: The master is RHEL 6 on a Sun/Oracle X4800, lots and
lots of fast cores and memory. I'm using Puppet Open Source. I'm using
passenger. I have no real modules other than some basic forge modules I've
installed to start out with.
[root@MASTER audit]# cd /etc/puppet/modules
[root@MASTER modules]# ls
apache concat epel firewall inifile passenger puppet puppetdb ruby
stdlib
I briefly disabled SELinux on the master, but saw no change in behavior.
I'm certain that the firewall is right because other agents have had no
problems. iptables IS enabled, however.
The master and the agent are on the same subnet, so I don't suspect a
network performance issue directly.
On Solaris, because the vendor supplied OpenSSL is antique and doesn't
include SHA256, we have built our own OpenSSL and our own Ruby using that
OpenSSL Library. Even though SPARC is a 64 bit architecture, Ruby seems to
default to a 32 bit build, so we built OpenSSL as 32 bit as well to match.
I've got an open question to the guy responsible for that to see how hard
it would be to try to build Ruby as 64 bit, that's likely a next test.
I have not yet run snoop on the communication to see what's going on the
network side, but as I say I don't really expect the network to be the
problem, between being on the same subnet and success on other systems with
higher clock speeds.
Any pointers to other possible causes or somewhere I can (even temporarily)
increase the timeout would be appreciated.
On Thursday, August 8, 2013 8:56:33 AM UTC-5, jcbollinger wrote:
>
>
>
> On Wednesday, August 7, 2013 11:46:06 AM UTC-5, Cesar Covarrubias wrote:
>>
>> I am already using Passenger. My master is still being minimally
>> utilized, as I'm just now beginning the deployment process. In terms of
>> specs, it is running 4 cores and 8GB of mem and 4GB of swap. During a run,
>> the total system usage is no more than 2GB and no swap. No network
>> congestion and I/O is low on the SAN which these VMs use.
>>
>> The odd thing is once the hosts get all the libs sync'd, performance is
>> fine on further changes. It's quite perplexing.
>>
>>
> To be certain that contention by multiple Puppet clients does not
> contribute to the issue, ensure that the problem still occurs when only one
> client attempts to sync at a time. If it does, then the issue probably has
> something to do with the pattern of communication between client and
> master, for that's the main thing that differs between an initial run and
> subsequent ones.
>
> During the initial plugin sync, the master delivers a moderately large
> number of small files to the client, whereas on subsequent runs it usually
> delivers only a catalog, and perhaps, later, 'source'd Files declared in
> your manifests. There may be a separate connection established between
> client and master for each synced file, and anything that might slow that
> down could contribute to the problem.
>
> For instance, if a firewall on client, master, or any device between makes
> it slow or unreliable to establish connections; if multiple clients are
> configured with the same IP number; if a router anywhere along the network
> path is marginal; if a leg of the path is wireless and subject to
> substantial radio interference; if any part of your network is suffering
> from a denial-of-service attack; etc. then probabilistically speaking, the
> effect would be much more noticeable when a successful transaction requires
> many connections and data transfers than when it requires few.
>
>
> John
>
>
--
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to puppet-users+unsubscr...@googlegroups.com.
To post to this group, send email to puppet-users@googlegroups.com.
Visit this group at http://groups.google.com/group/puppet-users.
For more options, visit https://groups.google.com/groups/opt_out.
