HSM process completes prematurely, no compression done

[Rich Graves]

I'm working on a Network Edition version 4.5.4 trial on RHEL4, 64-bit, default ext3 filesystems.

Through the admin web interface, I set up a secondary volume with default settings: 30 days, compress over 4096 bytes.

The default /opt/zimbra/store is part of the /opt filesystem, and I created /opt/zimbra/store3 in a separate lvm partition. It's called store3 because previously, I had confused myself too much with an /opt/zimbra/store2 in the /opt partition, never actually stored any messages there, and successfully deleted it.

I went into the HSM tab and fired off "Start HSM Session."

Messages appeared to move from the primary to the secondary fine. Disk usage went down on primary, up on secondary, and I see files being created in the hash tree.

However, aggregate disk space did not go down.

Poking through the drectory hierarchy, I see a lot of large messages in uncompressed form. I can't immediately find any that were in fact compressed.

I go into the GUI to check the progress. It says "completed" but the little green progress bar showed only about 75% done.

After watching vmstat/iostat/top for a while to convince myself that there wasn't still something going on in the background, I hit "Start HSM Session" again.

That returned immediately, reporting "completed" but the green progress bar is only 5% or so along. "Last HSM session was started on 4/6/2007 7:25 AM and ended on 4/6/2007 7:25 AM."

[Where] are there logs I can look at to find out why compression isn't being done?

[Rich Graves]

Attaching screenshot to clarify what I'm talking about.

The HSM process is correctly moving older blobs to the secondary volume.

But I have configured HSM to compress blobs over 4K (default). That compression is not happening.

The attached image suggests to me that Zimbra knows that the process is stopping after only half the work is done. It is "completed," but the progress bar is only at 50%.