[Box Backup] Error message (3/41)

[Ben Summers]

On 14 Feb 2004, at 06:04, Pascal Lalonde wrote:

> Feb 13 22:09:02 claymore bbstored[28253]: Login: Client ID 00000001,
> Read/Write
> Feb 13 22:20:36 claymore bbstored/hk[18538]: Starting housekeeping
> Feb 13 22:20:36 claymore bbstored/hk[18538]: Finished housekeeping
> Feb 13 22:25:29 claymore bbstored[28253]: in server child, exception
> Server (3/41) -- terminating child
> Feb 13 22:35:36 claymore bbstored/hk[18538]: Starting housekeeping
> Feb 13 22:35:37 claymore bbstored/hk[18538]: On housekeeping, sizes in
> store do not match calculated sizes, correcting
> Feb 13 22:35:37 claymore bbstored/hk[18538]: different (store,calc): 
> acc
> 0x00000001, used (586742,586951), old (79540,79745), deleted
> (9114,9114), dirs (3315,3315)
> Feb 13 22:35:37 claymore bbstored/hk[18538]: Finished housekeeping
>
>
> I guess this means the connection between the client and the server was
> interrupted, for example the ISP maybe have had a little problem.
> Then, the server noticed and inconsistency, and corrected it.

Yes, the client connection timed out, and the server process was 
terminated.

>
> I'm just trying to make sure. So everything should be in order right?
> This is normal behavior?

Absolutely.

The store has a "store info" file for each account, which gives 
information on last object IDs used, size of files, etc. It doesn't 
have to be absolutely up to date for things to work correctly, so is 
written lazily after every few transactions for performance reasons.

When an error occurs everything is cleaned up nicely, but the store 
info it isn't written back (as it might cause more errors) and so is 
then out of date. This doesn't matter, as the code is tolerant.

However, housekeeping needs accurate sizes. It scans everything anyway, 
so calculates the correct values and adjusts the store info file if 
they are wrong -- this is what you're seeing.

Ben