Re: Ontape restore hangs
Posted in 2008
Correction: I meant TAPEBLK below, not TAPESIZE.
"Doug Lawry" <lawry@nildram.co.uk> wrote in message
news:N5mdnfW0JLeCrkfVnZ2dnUVZ8v6dnZ2d@pipex.net...
> This turned out just to be slow performance rather than a hung ontape, and
> increasing TAPESIZE makes all the difference even when using backup images
> on disk. It may be peculiar to this disk array (Dell PowerVault MD3000).
>
> Astonishingly, with TAPESIZE 1024 (the largest suggested value I had seen)
> instead of 32 (the default), the restore completed in 55 minutes (about
> the same as "dd") instead of several hours, so I'll be making this change
> everywhere else!
>
> Top tip from this news group: use "onstat -D" to monitor "ontape"
> progress.
>
> "Doug Lawry" <lawry@nildram.co.uk> wrote in message
> news:XfOdnWpJ7rK1vkvVnZ2dnUVZ8vqdnZ2d@pipex.net...
>> Hi everyone.
>>
>> IDS 11.10.FC2 on SUSE 10.1. Migrating development instance to another
>> server and storage array (same OS, version and onconfig). 130GB
>> uncompressed ontape backup made on old system will not restore on the new
>> machine. Standard output stops after listing the dbspaces. Message log
>> stops on first checkpoint after starting physical restore. Process
>> running but nothing further happening 5 hours later.
>>
>> Tried both restore from directory device (ontape -p -d) and piped
>> (ontape -p -t STDIO). No different with "-r" instead of "-p". TAPEBLK 32,
>> TAPESIZE 0, ulimit unlimited. Shared memory checked clear beforehand
>> using "ipcs". No error messages other than the following which we always
>> get on startup (100 x 20MB logical logs):
>>
>> WARNING! Logical log layout may cause Dynamic Server to get into
>> a locked state. Recommended smallest logical log size
>> is 16 times maximum concurrent user threads.
>>
>> Worked around the problem using dd | rsh | dd and instance started fine
>> on first attempt, so dbspaces (or anything other than ontape) not the
>> problem.
>>
>> Any ideas? Cannot easily reproduce or run further diagnostics with ontape
>> running as the new system is now in use. Will raise with IBM unless you
>> guys come up with a theory.
>>
>> Regards,
>> Doug Lawry
>
>