Rainer -

I'll try and see if I can schedule a time to get those. Unfortunately these are 
the production servers, so it is difficult to test this. However I've just 
decided to schedule a time to take a traffic snapshot of traffic so I can 
replay it through our staging environment and see about simulating the issues 
I'm seeing.

I'll definitely setup the log.file settings for impstats. What you described 
sounds exactly like the issue I'm experiencing and makes complete sense now 
that you point it out. Thanks!

Radu -

I'll definitely make some of those changes to see how it impacts. I was able to 
confirm the ES cluster is able to handle far more traffic than is being sent to 
it. So I know it is the configuration I have in place on the Rsyslog side that 
I need to resolve. The queue.workerthreads should help in it's own. I've also 
increased the dequeuebatchsize to 1000 and we'll see how that impacts things.

I'll ad more information as I get it. (Rainer, I'll be sending the traffic to 
the Elasticsearch cluster again today. When I do I'll take a snapshot of the 
data being output impstats and post it for you to see.)

-- James

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Radu Gheorghe
Sent: Monday, October 14, 2013 2:37 AM
To: rsyslog-users
Subject: Re: [rsyslog] Queues And Max Sizes

2013/10/14 Rainer Gerhards <[email protected]>

> [...]

As you deal with a slow output, I don't think this will have much effect.
>
>
>
Right! It should help if you:
a) increase the batch size (20 is very little, especially if you have some fast 
servers). Somewhere between 100 and 1000 is a good starting point, although you 
might want to go higher if you have big boxes for ES.
b) increase the number of queue.workerthreads. Because, after sending a bulk 
rsyslog has to wait for the response from ES and parse it. So you'll get better 
throughput if you index logs on multiple threads (even more than you have CPUs, 
because of the waiting involved)

If you want to look at some more optimizations on the ES side, there are lots 
of them in my presentation Monitorama EU:
http://blog.sematext.com/2013/09/24/presentation-on-centralizing-logs/
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com/professional-services/
What's up with rsyslog? Follow https://twitter.com/rgerhards NOTE WELL: This is 
a PUBLIC mailing list, posts are ARCHIVED by a myriad of sites beyond our 
control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE THAT.
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com/professional-services/
What's up with rsyslog? Follow https://twitter.com/rgerhards
NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of 
sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE 
THAT.

Reply via email to