Re: RAC with OCFS on Win2K - archive logs destination

We're running Oracle 9.2.0.4 RAC in Win2k3 Server, 2-Node Cluster, attached to FC SAN. You didn't mention where your OCFS drives were (I assume on a SAN or other shared storage device). You also didn't mention if by "store" you were referring to the archive log destination that Oracle uses to write the archived redo logs to or to the area where you backup your archive logs to. I will assume that you meant the archive log destination (although, fyi, as Daniel said, if you're not backing up your archive logs to somewhere, start... today).

I would recommend using one (or more) of your shared (i.e. OCFS) drives to write the archive logs to. Oracle actually has some best practices guidelines for this, but RAC changes things a bit. During our implementation of RAC last year, we discovered that there can be problems if your nodes aren't all writing archive logs to the same place. I'm a bit fuzzy on the details now, but I believe that it had to do with recovery. Our original intention was to duplex the archive logs to both an OCFS drive and a local drive (per node), but the consulting company that helped with the project discovered some info
(possibly on metalink) that warned against this type of config on RAC
as it could somehow make recovery a bit more difficult if one node died completely. If I can find the info they gave us, I'll post it.

Given that, we decided against duplexing the logs entirely and we write them only to a single OCFS drive (R). Then, we run a script every 15 minutes that copies the archive logs from the R drive to the D drive of each node (this script actually triggers a log switch as well). This means that every 15 minutes, we have 3 copies of our logs. Note: If a log switch were to occur automatically within the 15 minute interval
(this happens rarely due to the way we've sized our redos), we'd only
have one working copy until the next interval hit, but we're relying on the redundancies that we've built into the SAN config to handle that. Then, every night, both nodes are backed up to an offsite location (we use IBM's Tivoli Storage Manager or TSM). In addition, we run a daily RMAN incremental backup and a weekly RMAN full backup that gets sent to TSM as well (there's also a daily RMAN full backup that's only for convenience purposes and does not get sent to TSM). This provides for a fairly redundant/robust backup solution for a facility our size (we manufacture nuclear fuel tubing).

Like I said, if I find that info on RAC and log locations, I'll post it.

Jesse Received on Tue Jan 18 2005 - 12:09:46 CST