ANALYZE/DISK/REPAIR

[Ewa Skotnicka]

Hi,
I'd love to hear someone advice if there are any restriction using ANALYZE/DISK/REPAIR ?
Few times happened that disk on which it started became unavailable, even if I tried to make a dir on it, process hung.
First time it happened for disk of application's logs (and more) and second time for system disk. We use OpenVMS 7.3-2.

Thanks in advance,
Ewa

[Ian Miller.]

$ ANAL/DISK/REPAIR does lock the disk volume during its operation so this can cause other accessors to wait.

[Ian Miller.]

DFU does have the /NOLOCK option but this can cause more errors to be reported than is necessary.

[Volker Halle]

Ewa,

ANAL/DISK without the /REPAIR option does not lock the volume, so it allows concurrent file system operations on the disk. The /LOCK_VOLUME switch can be used to have it lock the volume as well. This would provide a consistent view of the volume. Without /LOCK_VOLUME, you may get different error messages for multiple runs, if there is file activity on that disk.

ANAL/DISK/REPAIR does lock the volume and you can't overwrite this using /NOLOCK_VOLUME. All file system operations will be stalled until ANAL/DISK/REPAIR is finished, so you won't normally want to run this during normal operations hours.

Volker.

[Ewa Skotnicka]

But my analyze/disk/repair has been in state LEF, without any log information, from hours (it started on midnight).
I couldnâ t stop my process doing dir on that disk (at 8 am), similar was with other processes which wanted to log their data on that disk, they went to RWAST when I tried to stop them.
It looked for me like a deadlock.
I couldn't say where is the exact reason of problem (which process).
I think, but I haven't check it, that stopping my ANALYZE/DISK/REPAIR procedure would be possible and would eliminate problem.

[Volker Halle]

Ewa,

depending on other file system activity on the disk, it may be possible to get into a deadlock situation. You can analyze this using SDA (or AMDS) to follow the pending lock requests and the locks owning the desired resource in a non-compatible mode.

I have seen a situation like this, when using multiple ANAL/DISK on the same disk from various subprocesses.

To find out, if a process is hung (in LEF) waiting for a lock request, do this:

$ ANAL/SYS
SDA> SET PROC/IND=
SDA> SHOW PROC/LOCK

If the first lock displayed is NOT in Granted mode, the process is waiting on that lock. You then need to find out, which lock on the resource is blocking that lock:

SDA> SHOW RES/LOCK=

From that lock, you can find the process, which holds this lock and find out about the state of that process. This get's interesting in a cluster (locks are cluster-wide).

Here is a pointer to a TIMA article, which describes this kind of troubleshooting for a non-clustered system:

http://h18000.www1.hp.com/support/asktima/operating_systems/0092B6A5-7943C7A0-1C01E7.html

Volker.

[Volker Halle]

Ewa,

ANAL/DISK/LOCK_VOLUME or ANAL/DISK/REPAIR obtain the volume blocking lock in EXclusive mode:

F11B$b

This lock is only used to interlock all file activity on a volume. The lock is normally held in CR mode on any node, which has the disk mounted. XQP transactions (file system activity) occur under PW mode.

Volker.

[Ewa Skotnicka]

Volker,
Thank you for advice, I will try to use it next time, if I have a chance (it depends if it is system disk or not, if it is on production system).
For now it means for me that I cannot use ANALYZE/DISK/REPAIR in my housekeeping procedure, I should do it manualy, instead.

rgds
Ewa

[Volker Halle]

Ewa,

you can certainly use ANAL/DISK /without /REPAIR) in your regular housekeeping procedures to report, whether there are any major problems on the disks. Repairing those problems should be a manual action.

Volker.

[Ewa Skotnicka]

Volker, Ian,
Thank you very much for your advices.

Kind Regards,
Ewa

[comarow]

Just a note. More for previous versions.
Make sure bypass priv is enabled in case system access has been turned off.

For safety sake I always use it.

[Ewa Skotnicka]

Owner of analyze/disk/repair was fully privileged.
I have turned off repair option and I will use it manually.
I will try to trace reason of problem when it occures.

Thanks,
Ewa

[Sebastian Bazley]

If you run ANAL/DISK/REPAIR in a batch job, and the log file is on the disk being analysed, then you might see the behaviour you reported.