IDS Forum

Re: informix.LO_hdr_partn

Posted By: Marcus Haarmann
Date: Saturday, 12 April 2014, at 2:59 p.m.

In Response To: Re: informix.LO_hdr_partn (BOYCHO VELKOV)

Hi,

thank you for your input, sad that your problem has not really been solved yet.

You are right, two of our chunks have the maximum pages used.
And I suspect the answer lies there somewhere.

According to our logfiles, we had a temporary problem, which occured only twice, but when
the error occurred, no inserts did work (error -136, no more extents was thrown in the application).
After some tries (about 100, according to our logfiles), inserts succeeded again.
This situation was recurring after 20 minutes, when we decided to move traffic
to a new instance temporarily.

At the moment, the instance is running but we are not inserting data.
(we modified the application, to look in two databases, one active for write, one archive instance,
only for reading old data, which has some benefits, but I cannot be sure that archiving data from the
active instance in the archive will not provoke the error again. At least, customers would not be affected
at this stage.).

I have the idea that somebody deleted a record in a full chunk and the system
had a problem re-using the space, wanted to add a page, which was not possible due to
reached max pages. We are trying to reproduce that.
We have to find the problem, cause otherwise we would never know if the error
comes up again, which leads to ugly problems for our customers.
Maybe it takes a moment, but I do not want to copy 2.5TB data to a new instance, not knowing
if the same error could occur there, too.

Marcus Haarmann

----- Ursprüngliche Mail -----

Von: "BOYCHO VELKOV" <bobi@ndb.bg>
An: ids@iiug.org
Gesendet: Freitag, 11. April 2014 17:49:13
Betreff: Re: informix.LO_hdr_partn [32904]

Hi Marcus,

We had the same problem a few months ago with Informix 11.50.FC9 on HP-UX with
4 chunks in sbspace 1TB of data.

We did not have a time to open PMR(Case) because our system is 24x7. We find
workaround to rename the tables with BLOB-CLOB column and recreate new
tables in a new sbspace with big Metadata partition and AVG_LO_SIZE=2K(Lowest
value).

We are lucky, because developer agreed to change the application and recreate
the part of application with select in old and new tables.
But Im sure that after a year well have the same problem.

This is a part of your onstat cs output:
sblobdbs:'informix'.LO_hdr_partn 0x200005 16777215 0
sblobdbs:'informix'.LO_hdr_partn 0x200007 16777215 0
sblobdbs:'informix'.LO_hdr_partn 0x20000b 3310646 0
sblobdbs:'informix'.LO_hdr_partn 0x20000d 892734 2764897

In the 0x200005 and 0x200007 I suppose that Informix reach the pseudo-table
limit of 32GB.
For the others its strange. In our case the amount of all free pages were
0.
We had a lot of space for User Data and Meta Data, but in LO_hdr_parth we had
0 Free for all chunks.

Regards,

*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.

Messages In This Thread

• informix.LO_hdr_partn
  Marcus Haarmann -- Thursday, 10 April 2014, at 6:03 a.m.
  • Re: informix.LO_hdr_partn
    Marcus Haarmann -- Thursday, 10 April 2014, at 6:34 a.m.
  • Re: informix.LO_hdr_partn
    BOYCHO VELKOV -- Friday, 11 April 2014, at 11:49 a.m.
    • Re: informix.LO_hdr_partn
      Marcus Haarmann -- Saturday, 12 April 2014, at 2:59 p.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.