Thread: Cluster failure due to space

Cluster failure due to space

From
"Shea,Dan [CIS]"
Date:
I have had a cluster failure on a table.  It most likely was due to space.
I do not not have the error message anymore, but it was indicating that it
was most likely a space problem.  The partition was filled to 99%.  The
table is about 56 GB and what I believe to be the new table that it was
writing to looks to be 40 files of  1GB.

The problem is that it did not clean itself up properly.

The oids that I believe it was writing to  are still  there.
There are 56 files of 102724113.* and 40 files of 361716097.*.
A vacuum had indicated that there was around 16 GB of free space.
I can not find any reference to 361716097 in the pg_class table.
Am I going to have to manually delete the 361716097.* files myself?

Dan.

Re: Cluster failure due to space

From
Tom Lane
Date:
"Shea,Dan [CIS]" <Dan.Shea@ec.gc.ca> writes:
> The problem is that it did not clean itself up properly.

Hm.  It should have done so.  What were the exact filenames and sizes of
the not-deleted files?

> I can not find any reference to 361716097 in the pg_class table.

You are looking at pg_class.relfilenode, I hope, not pg_class.oid.

            regards, tom lane

Re: Cluster failure due to space

From
"Shea,Dan [CIS]"
Date:


"Shea,Dan [CIS]" <Dan.Shea@ec.gc.ca> writes:
>> The problem is that it did not clean itself up properly.

>Hm.  It should have done so.  What were the exact filenames and sizes of
>the not-deleted files?
361716097 to 361716097.39 are 1073741824 bytes.
361716097.40 is 186105856 bytes.

> I can not find any reference to 361716097 in the pg_class table.

>>You are looking at pg_class.relfilenode, I hope, not pg_class.oid.
Yes I am looking through pg_class.relfilenode.

Dan.