4 Replies Latest reply on May 12, 2016 2:35 AM by hchiorean

HikariCP pool is starting really slow

hchiorean May 11, 2016 7:34 AM

this is a cross post of a question which was posted as an article by x-out:

Hey guys,

our Modeshape server is starting really slow when he starts the HikariCP Pool.

Here is the snippet from our log file:

2016-05-11 08:05:30.243 INFO 14854 --- [WrapperSimpleAppMain] org.modeshape.jcr.JcrRepository          : ModeShape version 5.0.0.1-deg
2016-05-11 08:05:30.529 INFO 14854 --- [modeshape-start-repo-1-thread-1] com.zaxxer.hikari.HikariDataSource       : HikariCP pool HikariPool-0 is starting.
2016-05-11 09:02:14.583 INFO 14854 --- [WrapperSimpleAppMain] d.d.x.m.auth.SpringSecurityProvider      : [x] Successfully authenticated.

While it is starting, the server just uses 10% from one CPU.

In our database (postgresql v. 9.3) we have "just" ~600000 nodes.

The question is:

Do we need to configure something that the HikariCP thing is using the whole resources which are available.

Or can we do other performance-tuning to boost the start times?

Hopefully you can help us.

Regards.

1. Re: HikariCP pool is starting really slow

hchiorean May 11, 2016 7:36 AM (in response to hchiorean)

There are various options you can use to tune the HikariCP, but they aren't useful until you investigate what's causing the problem.
So I suggest first attaching a profiler and investigating (looking at CPU HotSpots, memory, stack traces etc) where the problem actually lies. Once you've identified the place, we can go from there.
Actions
2. Re: HikariCP pool is starting really slow

x-out May 11, 2016 9:44 AM (in response to hchiorean)

At start, we creating some indices.

For each entry in the indices, he ask the database, if the node is still present or not.
He is doing this one by one.

Is there any possibility to do this parallel?
Actions
3. Re: HikariCP pool is starting really slow

x-out May 12, 2016 1:45 AM (in response to x-out)

Hi,

i talked today with our chief architekt and he investigated this issue.
We identified an unused index, which ask every "nt:base"-nodes, that's stupid as hell.

So we deleted this index and restarted the modeshape-server. Now he is indexing again.

Looks like the problem is solved.
Actions
4. Re: HikariCP pool is starting really slow

hchiorean May 12, 2016 2:35 AM (in response to x-out)

glad you sorted it out.

In general, you can control the reindexing behavior via a "reindexing" section in the configuration. See Query and search - ModeShape 5 - Project Documentation Editor for more details.
Actions

Go to original post