I've found that 2.0.2 is rather "fat". When I upgraded from 1.x to 2.0.2 I 
had to add an additional Graylog instance and configure syslog-ng as a load 
balancer to process my load. Please check the resources that you're giving 
Graylog:

1) Are you giving Graylog at least 2gb of memory? Check your nodes list to 
see how much memory Graylog has been allocated.
2) Graylog can have long garbage collection delays with the default Java 8 
garbage collector. So if you want responsiveness it is best if you use the 
G1GC garbage collector.  In your GRAYLOG_SERVER_JAVA_OPTS in 
/etc/sysconfig/graylog-server add "-XX:+UseG1GC" to the options (as well as 
-Xmx20000m to give 2gb of memory to Graylog).
2) How many CPU cores are you giving Graylog? Is /etc/graylog/server.conf 
modified to allow Graylog to start up enough threads to take advantage of 
all your cores?
3) What is your CPU load in 'top'? If it's pegged out at 100%, then yeah, 
you know what the problem is.
4) What about your Elasticsearch instance(s)? Have they enough memory?
5) Also, you may need to tune your index strategy to match how many 
Elasticsearch instances you're running. For example, if you have four 
Elasticsearch instances, in your server.conf you may want to define your 
index strategy in server.conf with elasticsearch_shards = 2 / 
elasticsearch_replicas = 1 so that you're using all four instances for your 
indexes. That will make anything that does an index search literally run 
four times faster.

Basically, if you've used a specific commercial (sp) log indexing solution 
(lunk) then figure you'll need at least eight times as many CPU resources 
to process the same number of events via Graylog, and, of course, you'll 
need a well endowed Elasticsearch cluster to get decent performance. In my 
case, I have four ElasticSearch instances and two Graylog instances to 
process the same amount of data that was processed with a single instance 
of the commercial solution, luckily that's all running on my internal cloud 
with underutilized paid-for machines or Graylog would actually be more 
expensive than the commercial solution. My guess is that you're just not 
throwing enough resources at Graylog, but you can verify that by ssh'ing 
into the graylog server and running 'iostat -c 1' (it's in the 'sysstat' 
package in case that's not part of your standard package load). If %idle 
shows 10% or less on a consistent basis, you probably need more CPU. 

On Thursday, June 16, 2016 at 7:17:27 AM UTC-7, Marko Lerota wrote:
>
> Hi guys. I installed few 2.0.2 versions and the web interface is really 
> slow. 
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Graylog Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/graylog2/4abafb66-bdd0-4be9-888e-faa1cae498e6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to