Re: [U2] UV: Other ways of replicating?
Ken - the nightmare was in the time and effort to get it setup for 100+ accounts and 1000 users. Once done it did work, tho they did have the odd hickup. On 1/19/06, Ken Wallis <[EMAIL PROTECTED]> wrote: > > Symeon Breen wrote: > > > > I know some people who did the transaction logging and replay trick a > > few years ago on unidata before it got replication - turned into a bit > > of a nightmare for them, tho in the end it did the job. I think os > > level would be best. > > Then they can't have put the right amount of time and effort into setting it > up. DR via RFS archiving works fine. > > The problem with OS level replication is that it works at a block level and > at any point of time if you have an outage the replica database may look > fine at the OS level, but be physically corrupt as far as the database is > concerned. > > Database replication is best done be the database even if you have to bend > it out of shape to make it work! > > Cheers, > > Ken > --- > u2-users mailing list > u2-users@listserver.u2ug.org > To unsubscribe please visit http://listserver.u2ug.org/ --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
RE: [U2] UV: Other ways of replicating?
> Symeon Breen wrote: > > I know some people who did the transaction logging and replay trick a > few years ago on unidata before it got replication - turned into a bit > of a nightmare for them, tho in the end it did the job. I think os > level would be best. Then they can't have put the right amount of time and effort into setting it up. DR via RFS archiving works fine. The problem with OS level replication is that it works at a block level and at any point of time if you have an outage the replica database may look fine at the OS level, but be physically corrupt as far as the database is concerned. Database replication is best done be the database even if you have to bend it out of shape to make it work! Cheers, Ken --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
Re: [U2] UV: Other ways of replicating?
I know some people who did the transaction logging and replay trick a few years ago on unidata before it got replication - turned into a bit of a nightmare for them, tho in the end it did the job. I think os level would be best. On 1/13/06, John Hester <[EMAIL PROTECTED]> wrote: > Michael Doyle wrote: > > Recently, we've been looking at ways of developing a high availability > > cluster for our production UniVerse environment. Thanks to > > www.linux-ha.org we've been able to install two machines at different > > locations (cities twenty miles apart) and if the primary machine drops, > > users can log into the same IP address within a few seconds. The problem > > we have run into is replicating our data across the two nodes. > > We're replicating UV at the OS level on RH linux with Steeleye's > LifeKeeper software: > > http://www.steeleye.com > > The backup machine keeps an entire external raid set synched at the > block level. Should the primary machine fail or stop responding over > the network, the backup machine shuts off power to the primary machine, > assumes its IP address, mounts the filesystems on the raid set, and > starts all necessary services (one of which is UV). We experienced some > crashes due to a kernel bug early on, so we've tested this setup in a > production environment. We never experienced any UV file corruption > after a failover. > > Our machines are in the same datacenter though, with a gigabit ethernet > connection between them - not 20 miles apart like yours. I don't know > if you could do this over a WAN connection. The people at Steeleye > should be able to tell you though. > > -John > -- > John Hester > System & Network Administrator > Momentum Group Inc. > (949) 833-8886 x623 > http://memosamples.com > --- > u2-users mailing list > u2-users@listserver.u2ug.org > To unsubscribe please visit http://listserver.u2ug.org/ --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
Re: [U2] UV: Other ways of replicating?
Michael Doyle wrote: Recently, we've been looking at ways of developing a high availability cluster for our production UniVerse environment. Thanks to www.linux-ha.org we've been able to install two machines at different locations (cities twenty miles apart) and if the primary machine drops, users can log into the same IP address within a few seconds. The problem we have run into is replicating our data across the two nodes. We're replicating UV at the OS level on RH linux with Steeleye's LifeKeeper software: http://www.steeleye.com The backup machine keeps an entire external raid set synched at the block level. Should the primary machine fail or stop responding over the network, the backup machine shuts off power to the primary machine, assumes its IP address, mounts the filesystems on the raid set, and starts all necessary services (one of which is UV). We experienced some crashes due to a kernel bug early on, so we've tested this setup in a production environment. We never experienced any UV file corruption after a failover. Our machines are in the same datacenter though, with a gigabit ethernet connection between them - not 20 miles apart like yours. I don't know if you could do this over a WAN connection. The people at Steeleye should be able to tell you though. -John -- John Hester System & Network Administrator Momentum Group Inc. (949) 833-8886 x623 http://memosamples.com --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
RE: [U2] UV: Other ways of replicating?
Could you use UV replication for the hashed files, and rsync for the type 1 & 19 files? Depending on your use of those directory type "files", this might accommodate your needs (as long as the hashed files and directory "files" are not linked together with transaction bracketing, etc, and your application could handle them being out of sync a little. /Scott Ballinger Pareto Corporation Edmonds WA USA 206 713 6006 -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Michael Doyle Sent: Wednesday, January 11, 2006 6:58 AM To: u2-users@listserver.u2ug.org Subject: [U2] UV: Other ways of replicating? UniVerse users, Recently, we've been looking at ways of developing a high availability cluster for our production UniVerse environment. Thanks to www.linux-ha.org we've been able to install two machines at different locations (cities twenty miles apart) and if the primary machine drops, users can log into the same IP address within a few seconds. The problem we have run into is replicating our data across the two nodes. The best way would be database replication. Unfortunately, UniVerse can't replicate the type 1 and type 19 files, which we use extensively. Further, we cannot use ALTER.TABLE or CLEAR.FILE on replicated files. (These from the Administrators manual.) The second option we've looked at is some sort of disk replication. IBM recommends DRBD for use with DB2. Unfortunately, it seems that this would not provide referential integrity. The third option we looked at was a complicated mix of transaction logs and NFS mounts. Our standby box would be constantly updating itself off of the transaction logs of the primary. In the event of an outage, the standby comes up more or less updated and the dead primary can refresh itself off of the standby's logs when it comes back up. Does anyone have any experience getting around the shortcomings of UniVerse's replication? Thanks, Mike Doyle Unix Developer / Administrator --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/ --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
[U2] UV: Other ways of replicating?
UniVerse users, Recently, we've been looking at ways of developing a high availability cluster for our production UniVerse environment. Thanks to www.linux-ha.org we've been able to install two machines at different locations (cities twenty miles apart) and if the primary machine drops, users can log into the same IP address within a few seconds. The problem we have run into is replicating our data across the two nodes. The best way would be database replication. Unfortunately, UniVerse can't replicate the type 1 and type 19 files, which we use extensively. Further, we cannot use ALTER.TABLE or CLEAR.FILE on replicated files. (These from the Administrators manual.) The second option we've looked at is some sort of disk replication. IBM recommends DRBD for use with DB2. Unfortunately, it seems that this would not provide referential integrity. The third option we looked at was a complicated mix of transaction logs and NFS mounts. Our standby box would be constantly updating itself off of the transaction logs of the primary. In the event of an outage, the standby comes up more or less updated and the dead primary can refresh itself off of the standby's logs when it comes back up. Does anyone have any experience getting around the shortcomings of UniVerse's replication? Thanks, Mike Doyle Unix Developer / Administrator --- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/