vibinsathyan
asked on
Zimbra opensource backup
We have an opensource zimbra server with around 1000 users.
Right now using rsync, we copy the opt/zimbra folder to another server to keep as backup weekly.
but during failover, it was showing "BLOB Error" and most of the clients were not able to sync with the server.
so it was a failure.any ideas for backup in opensource version.
We tried many methods but was not sucessful.
Right now using rsync, we copy the opt/zimbra folder to another server to keep as backup weekly.
but during failover, it was showing "BLOB Error" and most of the clients were not able to sync with the server.
so it was a failure.any ideas for backup in opensource version.
We tried many methods but was not sucessful.
ASKER
Thank you for the response.But whether we have any workaround for "no such blob" error.
or any other simple backup method as im not expert in linux or zimbra.
or any other simple backup method as im not expert in linux or zimbra.
You might not be an expert now, but if you are maintaining a mail server for ~1000 mailboxes, it would be a good time to start learning...
There is no "simple backup method" for Zimbra. For a consistent backup, the files must match the database, this requires the database to be offline while both the databases and files are backed up, the DRBD method that I linked to is not strictly speaking a "backup" as it is not possible to restore to a point in time, it is closer to a network level RAID1 (mirror) of the disk(s) in the Zimbra server, but can provide you with HA capabilities.
If your storage has the capability of taking a snapshot (LVM/BTRFS/ZFS/SAN) then you could stop MySQL and the Zimbra services, take a snapshot, start MySQL and the Zimbra services, then mount the snapshot and rsync the contents of the snapshot as a backup to other storage.
If you are running Zimbra on your own hardware, another alternative would be to utilise a hypervisor with replication capabilities such as Microsoft Hyper-V server (the free edition) to replicate your Zimbra VM(s) between two or more hosts, with replication snapshots this can provide more than just a "failover".
Is your primary aim to have a backup method or a HA method ?
There is no "simple backup method" for Zimbra. For a consistent backup, the files must match the database, this requires the database to be offline while both the databases and files are backed up, the DRBD method that I linked to is not strictly speaking a "backup" as it is not possible to restore to a point in time, it is closer to a network level RAID1 (mirror) of the disk(s) in the Zimbra server, but can provide you with HA capabilities.
If your storage has the capability of taking a snapshot (LVM/BTRFS/ZFS/SAN) then you could stop MySQL and the Zimbra services, take a snapshot, start MySQL and the Zimbra services, then mount the snapshot and rsync the contents of the snapshot as a backup to other storage.
If you are running Zimbra on your own hardware, another alternative would be to utilise a hypervisor with replication capabilities such as Microsoft Hyper-V server (the free edition) to replicate your Zimbra VM(s) between two or more hosts, with replication snapshots this can provide more than just a "failover".
Is your primary aim to have a backup method or a HA method ?
ASKER
Primary aim is to have a backup method where backup server can be used for failover.
That is why we tried with 2 similar servers - one as back up.
What we were trying is to use rsync to copy /opt/zimbra folder to other similar server where zimbra is already installed.
But failover was not successful because of "no such blob" error while clients trying to sync (all use pop3).
Is it because the data was 3-4 days old and mailboxes were different or any only HA method will work for failover in minimal time?
That is why we tried with 2 similar servers - one as back up.
What we were trying is to use rsync to copy /opt/zimbra folder to other similar server where zimbra is already installed.
But failover was not successful because of "no such blob" error while clients trying to sync (all use pop3).
Is it because the data was 3-4 days old and mailboxes were different or any only HA method will work for failover in minimal time?
if your primary aim is to have a HA solution, I would suggest that you looked into the DRBD method I linked to previously
ASKER
Thanks a lot for your suggestion.Ill try try this.Hope everything will be OK.
Please help me if I get stuck in any part.
Thank you again and wishing a very happy new year ahead.
Please help me if I get stuck in any part.
Thank you again and wishing a very happy new year ahead.
ASKER
Dear,
I tried the method you mentioned, but I seen it is exactly the same replica of primary server in secondary.
My doubt is , say primary database or some service got corrupted and mail server is stopped, what will be the impact?
It will replicate to secondary and both will get same error right?
I tried the method you mentioned, but I seen it is exactly the same replica of primary server in secondary.
My doubt is , say primary database or some service got corrupted and mail server is stopped, what will be the impact?
It will replicate to secondary and both will get same error right?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
i know this thread is closed, but snapshoting the filesystem behind a running mysql server is very likely to produce a mess if the server is busy at that time. it should start up in innodb recovery mode 1 and loose a few seconds worth of queries, though. zimbra has it's own replication mechanisms and can be run in a cluster. you should either run an actual cluster or frequent offline backups.
@skullnobrains, this is precisely why I said that MySQL and the Zimbra services should be stopped at the time of the snapshot.
One alternative would be to use block level replication such as DRBD, such as https://imanudin.net/2015/03/24/how-to-install-configure-zimbra-high-availability-ha/