AnsweredAssumed Answered

Very long delay to login to the web client

Question asked by konsultex on Jun 29, 2008
2 days ago my Alfresco 2.9 B on Debian with MySql, JRE 1.5.0_xx (I believe xx = 11) and Tomcat (the Alfresco Linux installer) was being used by a group of about 20 people. The server has 1 gb ram and a 160 gb hd.

That repository has 1 space with a Java Scritp rule that is able to move a document placed into a special folder directly to its destination folder based on the name of the file. This was working fine. I don't know how big the content store was but my guess is that not more that 10 gb. Yesterday I had to copy about 80 gb into the repository into a folder that is not subject to any of the content rules above. I did this by connecting an external usb drive to the server and then using a very comfortable console ftp program called ncftp (in case anyone finds it of use, here it is: http://www.ncftp.com/). With it I copied all the folders and files unattended in about 10 hours. It was all working fine. My content store now has 96 gb (as reported by the operating system using "du -sh" in the alf_data directory).

Unfortunately that folder had very permissive settings and during the day some people decided to move the sub-folders around. At a certain point there were complaints that Alfresco was not working. Part of the problem was that the folders had been moved around and others did not know about it. But the other part of the problem is that the log showed many heal out of memory errors. I think it happened after some massive folder moves. I increased -Xmx to 700m which is close to what the server can handle without starting very heavy swapping. After I managed to have Alfresco working again I noticed a very heavy load on the system (close to 9) but I suppose that the content store was being reindexed. I left it like that overnight. In the morning I stopped Alfresco and MySql and restarted them. I now have a normal load on the system.

I also used MySql to examine the alf_audit_fac, alf_node and alf_node_properties tables because management wants to know who moved things around. I have 380,000 records in alf_audit_fact. It seems like a lot but I suppose that most installations using auditing for a long period of time exceed this easily.

What happens now is that when I try to login using the web client, the cpu use goes way up and stays that way for close to 1 hour. After that the login completes. So this has me worried because I need to migrate this Alfresco to a "bigger" server where I have 4 gb ram and 500 gb storage and where hopefully it will all work again. I am worried that the login problem may happen there again. I can login using ftp just fine. alfresco.log does not show anything about this problem.

What is Alfresco doing for about 1 hour when I try to login? What should I be looking at or trying to solve the problem?

What I need is to get back to normal web client logins.

Thanks.

Outcomes