[p4] Trying to understand the benefits of "Warning! We stronglyrecommend that you set up your system so that the journal iswritten to a filesystem other than the P4ROOT filesystem."
sweeney at ea.com
Mon Nov 20 02:38:48 PST 2006
Having access to the journal after a db failure allows for intra-day
recovery in the three disk case (disk 1: db files; disk 2: depot files;
disk 3: checkpoints and journal). You can recover from last night's
checkpoint, AND replay the day's journal to get complete point-in-time
recovery in the event of loss of the db files, rather than just recover
to last night's checkpoint.
quis custodiet ipsos custodes -- Juvenal VI, 347-8
> -----Original Message-----
> From: perforce-user-bounces at perforce.com
> [mailto:perforce-user-bounces at perforce.com] On Behalf Of John D'Albis
> Sent: 17 November 2006 21:06
> To: perforce-user at perforce.com
> Subject: [p4] Trying to understand the benefits of "Warning!
> We stronglyrecommend that you set up your system so that the
> journal iswritten to a filesystem other than the P4ROOT filesystem."
> Currently, we have the journal and depot files on the same physical
> disk. We don't follow how this change would actually help in a
> restore... It is repeated several times in the admin guide making it
> appear to be an imperative:
> The only use we can determine for having this file on a separate file
> system is that, if we were to lose only the non-journal disk - it will
> show which updates will not be included in the restore. What are we
> missing? Here is our logic:
> Assumptions: We do a checkpoint followed by a full backup every night
> at 3 am. If, in the case of any disk failure, we want to be able to
> restore to a known state as quickly as possible, with the least amount
> of work lost. Journal and checkpoint are on disk A, *.db files and
> Depot are on Disk B.
> Scenario a) disk A crashes, lose Journal and checkpoint.
> Recovery procedure: Install a new A disk - create new checkpoint file
> and start a new journal. No work is lost.
> Scenario b) disk B crashes, lose *.db and depot files.
> Recovery procedure: Install a new B disk - restore depot from backup
> AND recreate *.db files from checkpoint on Disk A. (While we
> also have
> a journal file and hence the ability to restore the metadata
> to a state
> that is newer than the depot, doing so would be fruitless as the depot
> files themselves would only be at the state of our last
> backup - 3 am).
> Work since 3 am is lost.
> Scenario c) disk A & B crashes - lose everything!
> Recovery procedure: Similar to scenario b. Work since 3 am is lost.
> Is there a scenario that we're missing?
> John D'Albis
> perforce-user mailing list - perforce-user at perforce.com
More information about the perforce-user