Possible data loss

dormando dormando at rydia.net
Thu Mar 27 15:39:14 UTC 2008


Uhh. Weird?

Do you have a lot of rows in file_to_replicate or file_to_delete or 
file_to_delete_later? It could be working through those?

Are you running a fsck?

If you connect to the trackers and run !watch are they spewing any 
errors? Double check your domain classes have the correct mindevcount?

Try strace'ing mogstored? :\

I'd just strace everything I can get my hands on at this point and watch 
for files being deleted, then backtrace. Are both of your trackers 
running mogilefsd v2 now?

-Dormando

vitaliy scherbakov wrote:
> Hello.
> 
> Sorry for bothering you once again :(
> 
> we have 2 hosts for data store
> 
> Statistics for devices...
>  device     host            files     status
>  ---------- ----------- ---------- ----------
>  dev1       host1          340099      alive
>  dev2       host2          766206      alive
>  ---------- ----------- ---------- ----------
> 
> we are upgrade mogilefsd to v2 and upgrade mogstored on host1 to v2 but 
> not upgrade mogstored on host2 (it is v1 still)
> we are warring about that number of files on dev1 decreasing each time 
> we check stats (at the beginning of migration to v2 there were more than 
> 500000 files), but as we can notice we didn't loss our content stays the 
> same (no loss)
> 
> in mogilefs db we have follow:
> 
> mysql> select count(*) from file;
> +----------+
> | count(*) |
> +----------+
> |    71820 | +----------+
> 
> mysql> select count(*) from file_on;
> +----------+
> | count(*) |
> +----------+
> |  1105921 | +----------+
> 
> Guys, can you explain how this is possible? and maybe we are doing 
> something wrong.
> 
> Thanks.
> 
> -- 
> Best regards,
> Vitaliy.



More information about the mogilefs mailing list