Обсуждение: Anything "time" critical on www.postgresql.org VM?
I'd like to setup hot failover for that VM, so that if the server it is running on goes down, it will auto-come back up on a second server. So far, the finest I can get the replication granularity is 15 minutes, so there is the potential of 15 minutes worth of 'data' to be lost on a crash ... As I believe that most (if not all?) of the stuff on that VM is pretty static as far as files, and there is no email over there, I can't think of any reason *not* to do this for that VM ... but, I wanted to check here to see if anyone knows of a reason that the potential lose of '15 minutes' worth of data is a bad thing? If not, I'll get that setup ... I don't believe that any of the other VMs can have this setup yet, until I can get the 'gap' reduced between replicas, since they are all dealing with CVS in some form or another, or email ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
Marc G. Fournier wrote: > > I'd like to setup hot failover for that VM, so that if the server it is > running on goes down, it will auto-come back up on a second server. So > far, the finest I can get the replication granularity is 15 minutes, so > there is the potential of 15 minutes worth of 'data' to be lost on a > crash ... Hi Marc, What kind of replication is being used? :) Regards and best wishes, Justin Clift
On Mon, 27 Sep 2004, Justin Clift wrote: > Marc G. Fournier wrote: >> >> I'd like to setup hot failover for that VM, so that if the server it is >> running on goes down, it will auto-come back up on a second server. So >> far, the finest I can get the replication granularity is 15 minutes, so >> there is the potential of 15 minutes worth of 'data' to be lost on a crash >> ... > > Hi Marc, > > What kind of replication is being used? rsync right now, unless you know of something better that works under FreeBSD? I'd love to find something more 'real time', but haven't been able to find anything that could be run on an existing server ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
> -----Original Message----- > From: pgsql-www-owner@postgresql.org > [mailto:pgsql-www-owner@postgresql.org] On Behalf Of Marc G. Fournier > Sent: 26 September 2004 20:30 > To: pgsql-www@postgresql.org > Subject: [pgsql-www] Anything "time" critical on > www.postgresql.org VM? > > > I'd like to setup hot failover for that VM, so that if the > server it is running on goes down, it will auto-come back up > on a second server. So far, the finest I can get the > replication granularity is 15 minutes, so there is the > potential of 15 minutes worth of 'data' to be lost on a crash ... > > As I believe that most (if not all?) of the stuff on that VM > is pretty static as far as files, and there is no email over > there, I can't think of any reason *not* to do this for that > VM ... but, I wanted to check here to see if anyone knows of > a reason that the potential lose of '15 minutes' > worth of data is a bad thing? > > If not, I'll get that setup ... > > I don't believe that any of the other VMs can have this setup > yet, until I can get the 'gap' reduced between replicas, > since they are all dealing with CVS in some form or another, > or email ... There's nothing time critical on there as far as the 'official' services go. Dunno if any of the users have any other tasks running that might be, but I doubt it. I say go for it! Regards, Dave
Marc G. Fournier wrote: > On Mon, 27 Sep 2004, Justin Clift wrote: <snip> >> Hi Marc, >> >> What kind of replication is being used? > > rsync right now, unless you know of something better that works under > FreeBSD? I'd love to find something more 'real time', but haven't been > able to find anything that could be run on an existing server ... Oh well, it sounds like it'll have to do. They use rsync here too for file based syncronisation. It get's bitched and moaned about, but "does the job". Regards and best wishes, Justin Clift
> rsync right now, unless you know of something better that works under > FreeBSD? I'd love to find something more 'real time', but haven't > been able to find anything that could be run on an existing server ... Doesn't FreeBSD have support for OpenAFS and/or CODA? ... John
On Mon, 27 Sep 2004, John Hansen wrote: >> rsync right now, unless you know of something better that works under >> FreeBSD? I'd love to find something more 'real time', but haven't >> been able to find anything that could be run on an existing server ... > > Doesn't FreeBSD have support for OpenAFS and/or CODA? k, I looked at Coda, and liked the concept ... but, it looks like something that, to implement, I'll have to setup before I put anything on the server, I can easily move a server to it ... and, from everything I read, I'm not 100% certain that it would even do what I wanted ... each time I thought I found the answer, reading a bit further seemed to negate it :( The thing is, its really simple ... two servers, each with a large file system (/vm) ... on that file system are subdirectories by company_id and then domain under that ... a domain is active only on one server, so the only thing being written to /vm/id/domain would be on one server, but /vm/otherid/domain (or even /vm/id/otherdomain) ... so I want changes to /vm/id/domain from ServerA 'replicated' to ServerB, and /vm/id/otherdomain from ServerB to ServerA (which is what I'm doing with rsync) ... I had looked at unison also, but it looked to have similar 'lag' restrictions as rsync does, and some limitations as to the kinds of files it could send back and forth ... Using rsync, and some mods that Andrea (oicu) helped me with to parrallelize it, I've been able to get granularity down to 5 minutes instead of 15, at least based on 4 VMs right now taking ~150 seconds total to keep in sync ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
Marc G. Fournier wrote: <snip> > Using rsync, and some mods that Andrea (oicu) helped me with to > parrallelize it, I've been able to get granularity down to 5 minutes > instead of 15, at least based on 4 VMs right now taking ~150 seconds > total to keep in sync ... Hmmm... I wonder if it'd be possible to use that PG backed filesystem someone created a while ago, then use PG replication (Slony, etc) to just replicate that? :) (Ok, prob not feasible, but it's an interesting thought) Regards and best wishes, Justin Clift
On Tue, 28 Sep 2004, Justin Clift wrote: > Marc G. Fournier wrote: > <snip> >> Using rsync, and some mods that Andrea (oicu) helped me with to >> parrallelize it, I've been able to get granularity down to 5 minutes >> instead of 15, at least based on 4 VMs right now taking ~150 seconds total >> to keep in sync ... > > Hmmm... I wonder if it'd be possible to use that PG backed filesystem someone > created a while ago, then use PG replication (Slony, etc) to just replicate > that? I'd thought about that, but for the kind of loads I put onto it, I don't know it would stand up ... it would definitely make failover alot easier, similar to doing NFS ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664