> Mark Murphywrites: > >Has anyone successfully configured checkpoint to use two >(or more) tape devices simultaneously? The old 'inquire_ingres' >newletter had an example of configuring the checkpoint >procedures to perform a "concurrent" checkpoint of all >database locations to a disk. This was fairly straightforward >because no modifications were required for the corresponding >rollforward. > >We must do our online checkpoints to tape, and to cut our checkpoint >elapsed times, we'd like to hear from anyone with experience >on this topic. Of course, mods to rollforward would be necessary. > >Environment: Ingres 6.4/04 (vax.vms), no journalling, >VMS 5.5-2 cluster. > >Thanks for any information... There is a paper on the NAIUA web site that I presented at Ingres World 95 that details this exact thing. Also, York and Associates has implemented such solutions for both VMS and UNIX at production customer sites. If you don't want to do it yourself, give us a call. Lastly, there is a bug in some versions of 6.4 (on VMS at least) that can cause problems during a parallel rollforward. Make sure you test your solution carefully. Cheers, -- |---------------------------------------------------| | Michael Leo York & Associates, Inc. | | mal@winternet.com (612) 921-8083 (voice) | | Minneapolis, MN, USA | |---------------------------------------------------| | NAIUA = North American Ingres Users Association | | http://www.naiua.org | | | | Ingres FAQ is at ftp.naiua.org in /pub/ingres | |---------------------------------------------------| From: scheib@olympia.mcis.washington.edu (Jeff Scheib) Organization: University of Washington Subject: Re: Checkpoint to 2+ tape devices concurrently Yah, we've juryrigged the ~ingres/files/cktmpl.def sufficiently to do all of: -checkpt to multiple disks -checkpt to multiple disks, simultaneously -checkpt to single tape(compressed and otherwise) -checkpt to multiple tapes -checkpt to two tape's simultaneously No "C" code needed, just general shell scripts. We have the "WST", "WSD", etc , call shell scripts which a) touches a token in /tmp b) tests for token file in /tmp c)if token file not found, start tar in background, else wait 1 minute and test again d)remove token file when tar completed. We sort of hard coded the location #'s into the logic so half of them ran one script (using /dev/rmt/tx1) and other half ran second script (using /dev/rmt/tx2). Be careful to add a "token testing" command in the "EE?" line of cktmpl.def which will "wait" till all tokens are removed prior to informing Ingres that checkpt has completed. Note: the lines in "BST" and "EET" are called only once during checkpt, and not once per location. And, get to know %A, %D, %N (chpt 15 of DBA guide) VERY well prior to ANY attempts/testing. And, save original cktmpl.def (I know, Duh!). These are the measures to which we are reduced when databases grow "larger". Hope you can go with this.
© William Yuan 2000
Email William