Slow send in web UI

[pbrunnen]

Hello all,

I have a strange one... I recently upgraded the server hardware to a new and much faster server for a customer. We are running zimbra (v6.0.7_GA_2473) on SLES 10 32-bit inside a VM on ESXi 4.1. Our original host was running ESXi 4.0.

Everything has been much faster across the board with the exception of sending emails. When you click send in the zimbra webui, there can be a delay of anywhere from nothing noticeable (<1s) to painful (>10s) before the message sends and the UI returns to normal.

In the attached txt file, you can see a time when there is a 10s delay between the soap SendMsgRequest, mailop and the SMTP Sending Message.

I don't see any errors being thrown... I verified our DNS servers...

Looking at the running processes, I see java and amavisd processes just after the send button is pressed. Then there is a period of dead time when nothing is happening... Finally another java and amavisd processes occurs that I can't seem to catch the full java command. I included in the attached file a snapshot of top from the exact moment the webui returns to normal and announces that the message was sent.

Any thoughts??
Thanks! -Cheers, Peter.

[kalba]

Hi,

Had somewhat of a similar problem on 6.0.9_GA but it turned out NOT to be Zimbra (I think). I was using the web ui with Firefox 3.6.13 and once I uninstalled Java Console 6.0.22 from Tools->Add-ons->Java Console 6.0.22 my slowness issue was solved.

Hope this helps and is relavent.

Tony

[pbrunnen]

Hi,

Well, I have proof now that the browser is unrelated. We had another stint of slowdowns (seem to be sporadic) with users onsite and offsite with and without company computers (multiple browsers and multiple platforms).

I did set my disk access shares to High for the zimbra vmdk in vmware...

I am really at a loss... nothing looks starved for resources...
Any other suggestions?

Thanks! -Cheers, Peter.

[pbrunnen]

Seems our problem is not just sending... All actions are responding slowly.

We looked into our host and we are not seeing any resource starving... but I did happen to capture on the short term graphing some wild cpu spikes. They are short and don't show up in our daily graphs.

There have been references to Java heap memory on the boards here...

Thoughts??