Engine crashes seconds after launching

[dangermouse]

Been working fine for some time, but just started crashing as soon as it launches. If I start it and then launch the console, they communicate for a moment before the engine dies. The engine crashed whether I launch the console or not.

 

The last thing I was aware the engine was doing was grooming a disk.

 

I have restored the Config80 files from a backup and also reinstalled the engine. Neither of these has helped.

 

I attached one of the crash logs.

 

Any ideas?

 

Many Thanks.

[Maser]

If the grooming "failed" and the engine crashed at some point prior to that, it's possible it's still trying to run a script accessing a corrupt/damaged media set and that's what is causing the crash.

 

It's possible your backup config80.dat file may still be trying to access that same media set.

 

What I might suggest would be moving the *catalog* files out of the folder you have them in and then restarting the engine. Scripts will not attempt to fully run if they can't find the catalog files.

 

That should (?) at least get you into the console to see what's going on.

 

Or try restoring a config80.dat file from an even older set. And delete the "config80.bak" file when you put the restored .dat file in place. I've seen instances where the .bak file was read because the engine believes the *.dat* file is hosed.

[dangermouse]

Thanks very much, removing the catalogues allowed it to run.

 

I identified the individual catalogue that was causing the problem. Then found even that catalogue could remain, I just had to remove the accompanying .rbc.log file of the same name. I presume this arose from the problem that occured while grooming.

 

Then within Retrospect, accessing that backup still caused a problem, but I then asked it to repair it, which it quickly did and all was fine.

 

The disk had only 9kb free, so obviously the grooming got nowhere. I presume I could groom again, but am not at all confident that this 10 hour plus process would work or could be trusted. I therefore now recycled that media.

 

Grooming appears to remain particularly fragile in Retrospect 8. And it takes so very much longer than the version 7 under Windows.

 

Fingers crossed this months update will fix this stuff!

[Maser]

You should always *rebuild* a catalog file after a failed groom. The rebuild can point out problems with the media set (in which case you have to figure out what to do...)

 

 

I didn't think it was possible to actually run a groom with only 9K (k? really?) free. I thought you had to have something like 1G free on the disk to be able to successfully run a groom (because the 630M .rdb files have to be able to be rewritten and need the swap space.)

[dangermouse]

The disk is set to groom automatically, so how much space is left is down to Retrospect. It began the groom in the middle of a large backup of over 20GB. So I think there was a fair bit before that.

 

I think the spare room requirement you mention is when the catalogue is on the disk? My catalogues are not on the backup disk, so there should be no problem?

 

Of course Retrospect should automatically see to this itself if there is an issue. If only I could find that manual, I must have misplaced it.

[Maser]

The 1G free space is if the catalog is *not* on the disk.

 

You need more disk space if the catalog is *on* the disk. (IIRC, you must have 5x the size of the catalog file free -- something like that).

 

So why it let you get to 9k before doing this, I don't know.

 

I think people get around this bug (?) by setting disk media set capacity so it doesn't use 100% of the disk space (99% or something is better...)