ASM Disk offline - State hung

Question

Hey in my ASM configuration I setup this
Disk Path ASM Name Failure Group
/dev/raw/raw4 Data1 FG1_Data [on SAN1]
/dev/raw/raw5 Data2 FG2_Data [on SAN2]
/dev/raw/raw6 Reco1 FG1_Reco [on SAN1]
/dev/raw/raw7 Reco2 FG2_Reco [on SAN2]
all good.
My I switched off SAN1. A query on v$asm_disk shows disk DATA1 & RECO1 as offline - hung state.
I did in the asm instance in sqlplus:
ALTER DISKGROUP DATA DROP DISK data1; -> success
ALTER DISKGROUP RECO DROP DISK reco1; -> success
ALTER DISKGROUP DATA ADD failgroup DATA3 '/dev/raw/raw4'; -> success
ALTER DISKGROUP RECO ADD failgroup RECO3 '/dev/raw/raw6'; -> success
The query on v$asm_disk stills show the DATA1 and RECO1 - but DATA3 and RECO3 as well.
Is it right, that the "old" entries will disappear, once the ASM instance is restarted ? I can´t do this at the moment.
In generell - why went the ASM disks to offline ?
SAN1 was switched off for about 5 minutes - and was then switched back on.
Is this expected behaviour ?

Answer 1

Hi Chris,
as long as the information is still partially valid, the devices will be displayed in v$asm_disk. Main reason is: Some Oracle processes still have handles open on these devices, and as long as they exist the devices can still be seen.
A Restart of ASM + database will definitely remove these entries (if they cannot be seen by the operating system).
Regarding your other question: Oracle can only detect if a device is going offline (errors in the database and asm alter log). If a device goes online again, Oracle does not check this (this would be a major overhead to rescan lets say every minute to see if a disk is back online).
Hence you have to reinclude the dropped/offlined disks yourself.
Depending on the version (10g or 11g) and depending on the diskgroup compatibility level for rdbms a disk can simply be returned into the diskgroup (11g - alter diskgroup <dg> set disk <disk> online) and resilvered or have to be reincluded into the ASM instance (no resilvering, but a complete rebalance 10g: alter diskgroup <dg> add disk <disk> to failgroup <fg> [force])
Sebastian