Nuxeo is slow

Hello everyone

Please, I need your help.

I am working to update our ECM platform to go from Nuxeo Nuxeo 5.8 to 7.10. For this, I cloned the production instance and I work on.

According to the documentation, I updated my Nuxeo with the latest hotfix (Nuxeo 5.8-HF39), and then I went from Nuxeo 5.8-HF39 to Nuxeo 6.0-HFHF00.

However, I noticed that Nuxeo is slower than before. The average response time has passed from 800 milliseconds to 2 seconds… I installed all HF nuxeo 6 (6.0-HF29) but it's still slow. I updated my hardware configuration by putting twice the RAM and twice the CPU but the result is the same…

I'have installed Nuxeo 7.10 TLS and Nuxeo is more slower…
I'have update my postgresql from 9.1 to 9.5 but it's not better…

You can see that I have a big latence in the attached screenshot

0 votes

3 answers

3367 views

ANSWER

You have to find what is slow before it can be improved. Is it the disk, the network, the database, the Java process… If the latter, then you'd have to use a profiling tool like visualvm to find where time is spent.
04/13/2016

Thanks for your help

Yes, that's what I'm trying to do.

The further I go, the more I think the problem is either nuxeo either the database.

Shibboleth is reached only once during authentication. Apache 2 is fast enough to process my request (I'm the only user logged).

It remains only Nuxeo and the database.

However, I do not think the problem comes from the database because it is the same as that of production.

The last point: Nuxeo … I note that Nuxeo 5.8 is faster than Nuxeo 6.0 I note that Nuxeo 6.0 is faster than Nuxeo 7.10 I see that 7.10 Nuxeo Nuxeo is faster than 8.1 etc …

Did you install an instance of nuxeo 7.10?

For information :

Here are the statistics of the document database generated by Nuxeo. Number of documents: 48928 Number of versions: 14275 Total Files: 95847 Total file size (KB): 116 685.877 Maximum Tree Depth: 16 the largest file size (KB): 1941 152.3 Record the heaviest 1752 children average speed reading documents (doc / s): 1 098.3

04/14/2016

You have to use a profiling tool like visualvm to find where time is spent. We can't debug a performance problem in the abstract.
04/14/2016



Hi,

For information, on the rendering layer, Nuxeo 6.0 contains an upgrade to JSF 2, which introduces some performance regressions which have been fixed on 8.2. These Optimizations should be available on 7.10 in the coming weeks (disabled by default).

1 votes



Hi Anahide,

Your comment is really interesting to us, as we're facing a similar issue for a migration from Nuxeo 5.6 to 7.10: after spending several days on the extensive use of profiling tools, all we could sum up was that the rendering of pages in general took longer than before.

Is there a JIRA where we can follow this issue/get more information about it?

Thanks

EDIT: Nevermind, found it at https://jira.nuxeo.com/browse/NXP-17690

05/03/2016


Hi

try creating some indexes in key tables such as hierarchy and all acls related. This will impact directly on performance. Also use elastic as provider and install it in an external node. Find more in https://doc.nuxeo.com/display/NXDOC/Configuring+the+Elasticsearch+Mapping

Regards

1 votes



Hello,

What is the difference between a default ElasticSearch and one that's I have create myself ?

04/14/2016


Thank you all,

I finally find the source of the problem. It was a hardware problem and not software. I had to move the VM to a new cluster to solve the problem.

bye :)

0 votes