Synchronization of 2 scalix server
Moderators: ScalixSupport, admin
-
edeles
Synchronization of 2 scalix server
We have 2 scalix servers. We want to make the other one as the backup. How do you synchronize the emails and user accounts between 2 scalix mail servers?Do we need to create the user accounts first in the backup server?
-
kanderson
I THINK you could install Scalix on the secondary server, and then do an rsync of /var/opt/scalix on a regular basis. That would be the cheap and crappy way, and it would have at least one problem. You'd want the same hostname and IP on both servers. The hostname you could likely use, even if it wouldn't really be correct. But you couldn't have two servers with the same IP. Having said that, you could have it as a cold standby, where you'd need to change the IP to bring it live. That should work...
Another way would be to use drdb to duplicate the server. This would allow writes on one server to be immediately replicated on the other. This will take some a seriously fast connection between the two servers.
The third and only supported way AFAIK, is to use a SAN, where both servers are connected to it. Since both servers are connected to the same data storage, replication is included by design.
Another way would be to use drdb to duplicate the server. This would allow writes on one server to be immediately replicated on the other. This will take some a seriously fast connection between the two servers.
The third and only supported way AFAIK, is to use a SAN, where both servers are connected to it. Since both servers are connected to the same data storage, replication is included by design.
-
jhamill
- Posts: 66
- Joined: Thu Dec 01, 2005 5:25 pm
kanderson wrote:I THINK you could install Scalix on the secondary server, and then do an rsync of /var/opt/scalix on a regular basis. That would be the cheap and crappy way, and it would have at least one problem. You'd want the same hostname and IP on both servers. The hostname you could likely use, even if it wouldn't really be correct. But you couldn't have two servers with the same IP. Having said that, you could have it as a cold standby, where you'd need to change the IP to bring it live. That should work...
Another way would be to use drdb to duplicate the server. This would allow writes on one server to be immediately replicated on the other. This will take some a seriously fast connection between the two servers.
The third and only supported way AFAIK, is to use a SAN, where both servers are connected to it. Since both servers are connected to the same data storage, replication is included by design.
this san or iscsi option is quite interesting to us. We are currently using the small business version of scalix (50 outlook users) but would really like high availability access to our emails. How would we go about clustering a scalix server, does it support such a thing?
-
kanderson
It does support HA, for sure, and has for a while. Unofficially, it's worked for MUCH longer.
You'll want to start here. Most people get it all working after reading this guide, and asking back here if they hit a snag.
http://www.scalix.com/wiki/index.php?ti ... -HACLUSTER
Note that this is for 11.1 and beyond.
Kev.
You'll want to start here. Most people get it all working after reading this guide, and asking back here if they hit a snag.
http://www.scalix.com/wiki/index.php?ti ... -HACLUSTER
Note that this is for 11.1 and beyond.
Kev.
-
Valerion
- Scalix Star
- Posts: 2730
- Joined: Thu Feb 26, 2004 7:40 am
- Location: Johannesburg, South Africa
- Contact:
You have a SBE Scalix install. Unfortunately according to http://www.scalix.com/enterprise/editions/compare.php high-availability is only available to the Enterprise Edition.
-
kanderson
-
Derek
- Posts: 169
- Joined: Fri Mar 24, 2006 4:53 pm
- Contact:
-
kanderson
If you're willing to accept some data loss, and down time, then you could "cluster" without that.
Basically, you'd mirror the data between two servers. When server 1 went down, anything not mirrored would be lost. On a large mailserver, I've seen the setup for an rsync take over an hour, so this could be a significant amount of data.
Second, since the rsync will not have completed, your mailstore will be inconsistent, so you may well need to run omscan against it before some users would even be able to get their email. This may not always be the case, but it may be, and it would be prudent to expect that it would be in the event of a failure, because this plan is designed to incur some replication failure. Not planning to handle that failure also would be stupid.
Since omscan itself can take a significant amount of time, and may need to be ran more than once, you will encounter some downtime here.
Having said that, this can work, and if nothing else, it would allow you to do a backup on a server that isn't live. (rsync it, and then omshut for as long as a backup takes). Also note that the bigger your mailstore, (the more users you have, and the more data they keep in there...) the worse the problems you'll encounter with this plan.
I'd recommend against it, personally. I suspect that the downtime of your mailserver will cost more than the additional cost of a license that will allow you to set this up correctly. And, as I already pointed out, since this will mean more downtime as you add users, it will continue to add value as you grow. So my blunt recommendation is to buy the license, and do it right. Trust me, if you do have a catastrophic failure, those few hundred dollars you've saved will evaporate quickly as your users are yelling at you to get the server back up, and you're emailing Scalix support to open an incident.
And even there, I'd talk to Scalix about that too, because it's possible that they'd just say that what you're doing is in violation of your license, so before they start to help you, you need to 1) purchase the correct licenses, and 2) pay for the support incident. Meaning you've created a headache, and also not saved yourself any money.
But if you want to ignore this advice, then it is possible to do, yes.
Kev.
Kev.
Basically, you'd mirror the data between two servers. When server 1 went down, anything not mirrored would be lost. On a large mailserver, I've seen the setup for an rsync take over an hour, so this could be a significant amount of data.
Second, since the rsync will not have completed, your mailstore will be inconsistent, so you may well need to run omscan against it before some users would even be able to get their email. This may not always be the case, but it may be, and it would be prudent to expect that it would be in the event of a failure, because this plan is designed to incur some replication failure. Not planning to handle that failure also would be stupid.
Since omscan itself can take a significant amount of time, and may need to be ran more than once, you will encounter some downtime here.
Having said that, this can work, and if nothing else, it would allow you to do a backup on a server that isn't live. (rsync it, and then omshut for as long as a backup takes). Also note that the bigger your mailstore, (the more users you have, and the more data they keep in there...) the worse the problems you'll encounter with this plan.
I'd recommend against it, personally. I suspect that the downtime of your mailserver will cost more than the additional cost of a license that will allow you to set this up correctly. And, as I already pointed out, since this will mean more downtime as you add users, it will continue to add value as you grow. So my blunt recommendation is to buy the license, and do it right. Trust me, if you do have a catastrophic failure, those few hundred dollars you've saved will evaporate quickly as your users are yelling at you to get the server back up, and you're emailing Scalix support to open an incident.
And even there, I'd talk to Scalix about that too, because it's possible that they'd just say that what you're doing is in violation of your license, so before they start to help you, you need to 1) purchase the correct licenses, and 2) pay for the support incident. Meaning you've created a headache, and also not saved yourself any money.
But if you want to ignore this advice, then it is possible to do, yes.
Kev.
Kev.
-
jhamill
- Posts: 66
- Joined: Thu Dec 01, 2005 5:25 pm
I think you are right. mirroring our physical scalix server's volume onto an iScsi volume, which could quickly be "re-attached" to a new server in the event of the first server falling over sounds like it might be the way to go.
Also there shouldn't really be an data loss with this scenario although some downtime, it could probably be measured in the minutes.
Also there shouldn't really be an data loss with this scenario although some downtime, it could probably be measured in the minutes.
kanderson wrote:If you're willing to accept some data loss, and down time, then you could "cluster" without that.
Basically, you'd mirror the data between two servers. When server 1 went down, anything not mirrored would be lost. On a large mailserver, I've seen the setup for an rsync take over an hour, so this could be a significant amount of data.
Kev.
Who is online
Users browsing this forum: No registered users and 5 guests