cancel
Showing results for 
Search instead for 
Did you mean: 

High CPU utilization

raju_myadam
Champ in-the-making
Champ in-the-making
Hello,

We have released an application into production recently. The application is built on top of community version of Alfresco 2.1. We are seeing high CPU utilization around 75% even when the system is idle with no user activity.
We are using following features of Alfresco -
Alfresco Document management
CIFS
FTP
Web services

Here are the environment details -
Hardware - Intel x86
Open Solaris 10
JDK 1.5
VMWare

Any advice on this behavior would be greatly appreciated.

Regards,
Raju
16 REPLIES 16

mrogers
Star Contributor
Star Contributor
It may be worth tuning your environment, have you given the JVM enough RAM for example?  

What are you using VMWare for?

Does your server box have anything else on it?  Is your server running well without Alfresco?

It could also be async processes such as the full text indexers.   How long has your production system been stable?

raju_myadam
Champ in-the-making
Champ in-the-making
Thanks for your response.

JVM min and max settings are 512 MB. The total RAM is 1 GB dual CPU.

It's a blade server, multiple partitions are running and managed by VM Ware.
It's been 3 weeks since we deployed the application in production. We are seeing high CPU utilization since then. On the same blade server we have jboss running which is perfectly fine in terms of CPU.

Regards,
Raju

nadaoneal
Champ in-the-making
Champ in-the-making
We were having a very similar problem, but worse, so that we were getting "java.lang.OutOfMemoryError: PermGen space" in the error log and CPU utilization was going past 90%. The solution in our case was to add a MaxPermSize as well to our JAVA_OPTS…
"-Xms256m -Xmx512m -Xss96k -XX:MaxPermSize=128m -server"
… hope that helps.

raju_myadam
Champ in-the-making
Champ in-the-making
Thanks for your suggestion.

We have identified the cause for the high CPU utilization. The problem was due to CIFS access. Whenever we access the space through CIFS protocol we have seen the CPU utilization spike. We had a batch process runs frequently and access the space through CIFS protocol. We have to re-engineer the solution to reduce the usage of accessing the spaces through CIFS protocol.

Regards,
Raju

beren
Champ in-the-making
Champ in-the-making
If it is a blade server - what type of drive system are you using?

Many blade systems come with low-end SCSI and even IDE/SATA drive systems rather than decent NAS/SAN storage. Local low-end storage on blades can really drive up the CPU since there is no advanced controller card or caching services.

beren

zomurn
Champ in-the-making
Champ in-the-making
I have the same problem on a dedicated production server for alfresco.
CPU Load is at 400% and vary between 100% and 400% on a quad core Xeon with 4Go RAM.
Alfresco startup command line is :

/home/oracle/jdk1.5.0_16/bin/java -Xms512m -Xmx1024m -server

I think this is a normal behavior, and lucene indexes are computed in background.
If it is the case, my question is how can I schedule this process the night ? Which file to modify ?

Thanks

lasse_mammen
Champ in-the-making
Champ in-the-making
I don't think this is due to indexing. When using WEBDAV, the cpu load is much lower and it feel way more responsive as well.

Would be nice to have some kind of solution and/or answer to this.

amathewk
Champ in-the-making
Champ in-the-making
Hi,

I also have the same problem. I am running on a shared VPS on CentOS.

There are only 3 word documents and one pdf loaded into a alfresco site. I am the only user accessing the system. The cpu usage starts off fine and then spikes to 100% after a while.

My site and vps account is responsive but the admins keep warning me about my excessive usage. Please let me if you have any ideas.

thanks

Anil

zomurn
Champ in-the-making
Champ in-the-making
My problem comes and comes again…I'am almost sure (very) it is a problem of configuration : that's lucene backup task ? I imagine a lucene indexing task that use 100% CPU.
I really need help to this it regularly happens in production or whatever.

Thanks