Re: Un-attended Large Archives (fwd)
Posted in 1993
In <1993May6.164433.27022@informix.com> davek@newjersey.informix.com (Dave Kosenko) writes: >Chad Larson writes: [stuff deleted] >|> >|> In fact, the whole purpose of this subtrafuge is to keep Informix from >|> seeing end-of-media, to avoid the global lock imposed during the media >|> change period. >I'm afraid this last statement is incorrect. There is no "global lock imposed" [stuff deleted] >In this situation, your system can be "frozen" until the tape is changed so >the archive can continue and get the physlog pages which allows the checkpoint >to complete which allows users to continue (whew!). >You imply that, by design, the system is intentionally locked during media >changes and this is incorrect. In fact, this problem can generally be I think you're being a bit pedantic here, Dave. The lockup effect may not be implemented as a "Global Lock" but nevertheless the effect is that of a global lock. >avoided by choosing a checkpoint frequency longer than the default 5 minutes >(there is little reason to have such frequent checkpoints, in my opinion), >better yet, choosing a checkpoint interval greater than the amount of time >it would take to change the tape. While a "freeze" is still possible >(should your checkpoint happen to fall while a tape is being changed anyway), >it becomes less likely than with frequent checkpoints. >I usually suggest letting the size of your physical log determine checkpoint >frequency (i.e. set CKPTINTVL to a large number). But what does this do to your Fast Recovery times??? We are trying to implement highly available systems here. We frequently get asked to provide recovery times after powerout of the order of 3 minutes. Since the system requires longer than this to boot and fsck, every second I can save on Fast Recovery is valuable, thus it is tempting to *increase* checkpoint frequency to the 30 - 60 second range. Incidentally do you (or anybody else) have any suggestions for how to predict average Fast Recovery time? I get the impression that this tape change/lockup problem is perceived as relatively low priority by Informix, whereas it is a *BIG ISSUE* for us, and I'm sure for any large site. What is the point of having an Online Archive facility if using it freezes your system as soon as you go beyond one tape unless your tape operator a) is on his/her toes and b) has nothing else to do??? I hope someone in Informix will tell me that my impression is wrong! graeme -- Disclaimer: The author's opinions are his own, and not necessarily those of Pyramid Technology Ltd. or Pyramid Technology Inc. --------------------------------------------------------------------------- -m------- Graeme Sargent Voice: +44 (0)252 373035 ---mmm----- Senior Database Consultant Fax : +44 (0)252 373135 -----mmmmm--- Pyramid Technology Ltd. Telex: Tell who??? -------mmmmmmm- Farnborough, Hants GU14 7PL Email: graeme@uk.pyramid.com --------------------------------------------------------------------------- We have the technology. The tricky bit is learning how to use it. operator