After full restore of Plesk server mailboxes being deleted, DNS pointing to old data

[eugenevdm]

We have six Plesk servers in our Expand hierachy:

2 x Plesk Linux Servers
2 x Plesk Windows Servers
1 x Central Mail Server
1 x Central DNS Server

We have to restore 1 x Plesk Linux server. We did a clean installation of Plesk and then restored the backup file. Now we have numerous major problems.

1. The first and biggest problem is some domains (with their mailboxes!) are just dissappearing off the Central Mail server. We have lost numerous mail accounts and domains and we are still loosing some more. It seems Plesk Expand is trying to sync and in the process deleting old clients and mailboxes off the Central Mail server.

2. Some clients had updated their DNS in the Expand hierarchy. Now these clients domain DNS records are pointing back to the old server, the updated records are gone. We cannot determine how wide this problem is but we are having multiple calls about it.

3. Expand client relations with Plesk client relations on the restored server is missing. There are over 300 clients to assign and we cannot do this by hand. We need an automated way to do it.

We need some serious help and we are always logging a paid for ticket about this. Please if anyone has advice give it now.

[eugenevdm]

Whilst following this article

http://kb.parallels.com/en/5861

at the step where you de-assign the Plesk mail server, we are getting this error:

Lock wait timeout exceeded; try restarting transaction. Error query: DELETE FROM cmail_attempt WHERE object_id=7236</errtext>

What can we do?

[eugenevdm]

It seems the culprit here is that the cmail_attempt that is almost 18GB.

It seems that cmail_attempt grows with millions of records every month.

This is causing the entire machine to become slow, or at least, Central Mail is screwed up.

I had to split the Expand database from ibdata1 and restore from backup. I did so by backing up and restoring and then using innodb_file_per_table=1. For good measure I also added innodb_lock_wait_timeout = 500.

But the table is still very big, down to 14GB now, but it's very difficult to delete records. It keeps on saying lock timeout or just hangs. I have Parallels support helping me now.