Re: Unix or hardware mirroring vs. Informix Online mirroring
In <23uhu6$mj3@access1.speedway.net> tomlowe@speedway.net (Tom Lowe) writes: >Is either one better than the other? Any comments or suggestions? Here's one that I suspect might set the cat among the pigeons! It seems to me that the following recent problem (which admittedly strikes me as an opsys problem not an OnLine problem) would not have arisen if opsys level mirroring had been used rather than OnLine mirroring: In <23tmrtINNn28@emory.mathcs.emory.edu> carlton@alexis.attmail.com writes: |8.5.93 | |I've got a question for someone pretty familiar with OnLine internals and |page allocation. | |I'm running OnLine 5.0 UC1 on an AT&T StarServer E with Sys V Rel 4.0 2.1, |and have several scsi controllers in the machine. All my dbspaces use |"raw" disk slices and are mirrored across to a drive on another controller. |I noticed the following message on the console of this machine --apparently |from the O/S-- just a little while ago: | | WARNING: SD01 I/O Error ha0tc5 Unit 0 Error 6dd03002 block 0x00051386 | Count 16384 Sense key 0x0000003 Extended Sense 0x0000011 Op code | 0x0000028 | | WARNING: SD01 Data from block 0x00051386 is not readable. Data from this | block has been lost. An alternative block has been assigned. | |I checked my engine log and there weren't any errors reported nor were any |of the dbspaces on that drive taken off-line by the engine. | |My questions are as follows: | |1. If in fact there was a bad block, why did the O/S catch it and not the | engine? Why didn't the engine report back a read or write failure? | |2. If the data was in fact lost as the message indicates, why didn't the | engine take the drive off-line and kick the mirror into production? | |3. If an alternate block was assigned, has the dbspace allocation structures | been updated to reflect the new free list etc. so the engine doesn't | try to write over a now used block? | |4. Have similar changes occured within the mirror? Is it still an exact | duplicate bitmap of the production drive or are the two now out of | sink because the production has data stored in one (relative) area of the | dbspace and the mirror store in another (relative) area of the mirror | dbspace? | |Any help appreciated. | |Carlton | |============================================================================ |Carlton Doe |Spectrum Field Services |v (801) 461-8241 |f (801) 461-8222 |AT&T Mail: attmail!alexis!carlton |Internet: cdoe@sunset.utah.edu OR att!attmail!alexis!carlton@uunet.UU.NET What I believe has happened is that the opsys has updated one side of OnLines mirror but didn't know how to tell OnLine to resync its mirrors (indeed it didn't know it *was* a mirror!). My HO is that opsys level mirroring (provided it works reliably) cannot be worse than database level mirroring (assuming it also works reliably) but could *potentially* be better. The database vendors should continue to provide database level mirroring however, until there are no environments left which do not support (reliable) opsys level mirroring. 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@pyra.co.uk --------------------------------------------------------------------------- We have the technology. The tricky bit is learning how to use it.