How to repliacate device failure ?

Hi All, Im doing some test on mirroring and wanted to replicate a primary chunk device failure. Is there any way to test this ? Well here is the scenario I have. Ive got Informix 9.4 (linux version). Two HDs - 1 primary chunk device and other is mirror chunk device. I created a test database with mirror turned on and took a level 0 backup (the doc says the mirroring start only after level 0 if u have mirroring on for logical logs). The primary chunk file for test database is testdbs. I filled testdbs on the primary device with bad blocks (thru e2fsck with -L option), I thot this wld bring the chunk down and mark it as bad device. But the test never succeeded. Is there any tools/methods to test the fail over mechanism of mirroring coz this is going to be implemented on a critical production server. Thank you in advance. -- Anil Menon MIS Team. I'll try till I die or die trying.

Hi, you will see messages in the message log file only when you use the system, i.e. you need to have ongoing work (transactions) on the system that use the dbspace (chunk) that you will be corrupting. I'm not sure what "e2fsck -L" is doing ... Why don't you use the "dd" utility ? It works excellent for this kind of thing, e.g.: dd if=/dev/zero of=<your chunk path> bs=2048 skip=3 count=12 This would write all-zeroes over the chunk, with a block size of 2kB (the page size), it would start at the fourth page and write 12 pages, thus corrupting pages 3 (start counting with 0) through 14 of the chunk ... In case you'd have an offset specified for your chunk, then you need to account for that with the "skip=" parameter ... Once you get the idea, you can easily play with the parameters "skip=" and "count=" and do this "work" repeatedly ... Doing such stuff on a running system (i.e. with transactions ongoing) should give you the opportunity to see the mirroring at work. It will also let you simulate "repairing" the disk and getting things back to normal - without disrupting IDS operations. Regards, Martin -- Martin Fuerderer IBM Informix Development Munich Data Management Solutions forum.subscriber@iiug.org wrote on 11.05.2004 08:19:00: > Hi All, > > Im doing some test on mirroring and wanted to replicate a primary chunk > device failure. Is there any way to test this ? > > Well here is the scenario I have. Ive got Informix 9.4 (linux version). > Two HDs - 1 primary chunk device and other is mirror chunk device. I > created a test database with mirror turned on and took a level 0 backup > (the doc says the mirroring start only after level 0 if u have mirroring > on for logical logs). The primary chunk file for test database is > testdbs. I filled testdbs on the primary device with bad blocks (thru > e2fsck with -L option), I thot this wld bring the chunk down and mark > it as bad device. But the test never succeeded. > > Is there any tools/methods to test the fail over mechanism of mirroring > coz this is going to be implemented on a critical production server. > > Thank you in advance. > > -- > Anil Menon > MIS Team. > > I'll try till I die or die trying.

Hi, Is there any tools/methods to do the same on WinK? Thank you in advance. Paola Quoting "Martin Fuer...." <MARTINFU@de.ibm.com>: > Hi, > > you will see messages in the message log file only when you use the > system, > i.e. you need to have ongoing work (transactions) on the system that use > the > dbspace (chunk) that you will be corrupting. > > I'm not sure what "e2fsck -L" is doing ... > Why don't you use the "dd" utility ? It works excellent for this kind of > thing, e.g.: > dd if=/dev/zero of=<your chunk path> bs=2048 skip=3 count=12 > This would write all-zeroes over the chunk, with a block size of 2kB > (the page size), it would start at the fourth page and write 12 pages, > thus > corrupting pages 3 (start counting with 0) through 14 of the chunk ... > In case you'd have an offset specified for your chunk, then you need to > account > for that with the "skip=" parameter ... > > Once you get the idea, you can easily play with the parameters "skip=" > and "count=" and do this "work" repeatedly ... > > Doing such stuff on a running system (i.e. with transactions ongoing) > should > give you the opportunity to see the mirroring at work. It will also let > you > simulate "repairing" the disk and getting things back to normal - without > disrupting IDS operations. > > Regards, > Martin > -- > Martin Fuerderer > IBM Informix Development Munich > Data Management Solutions > > forum.subscriber@iiug.org wrote on 11.05.2004 08:19:00: > > > Hi All, > > > > Im doing some test on mirroring and wanted to replicate a primary chunk > > device failure. Is there any way to test this ? > > > > Well here is the scenario I have. Ive got Informix 9.4 (linux version). > > Two HDs - 1 primary chunk device and other is mirror chunk device. I > > created a test database with mirror turned on and took a level 0 backup > > > (the doc says the mirroring start only after level 0 if u have mirroring > > > on for logical logs). The primary chunk file for test database is > > testdbs. I filled testdbs on the primary device with bad blocks (thru > > e2fsck with -L option), I thot this wld bring the chunk down and mark > > it as bad device. But the test never succeeded. > > > > Is there any tools/methods to test the fail over mechanism of mirroring > > coz this is going to be implemented on a critical production server. > > > > Thank you in advance. > > > > -- > > Anil Menon > > MIS Team. > > > > I'll try till I die or die trying. > > > >