DSXchange

Hi all,
I am using a server job that look like that : seq_file ->transformer -> transformer -> link_collector -> sort -> transformer -> aggregator ->transformer -> seq_file
The job uses inter process (with 512k buffer size)
When running on large amount of data I get an error:
ds_ipcput() - timeout waiting for mutex"
any suggestions?

Suggest a search is in order.

chulett wrote:Suggest a search is in order.

I searched the forum but haven't found any good suggestion for this problem
I'm looking for information about tuning my machine to prevent errors like this to happen

Reduce the total load and in particular don't do anything that needs to wait for an unexpected amount of time. Had you searched for SPINTRIES you would have found more useful tuning information.

Manageg to resolve the error by reducing the buffer size of the job.

I don't realy know why it helped... any idea?

One possible reason is physical memory. Large buffers used so much memory that the system began paging / swapping. This slowed user level processes down so much that your error occurred. This is only a guess, without additional data and your monitoring nobody can know for certain.

That was my thought as well.

Any wait can cause it. Even waiting for DB2 to generate all the warnings it needs to before rolling back a transaction, if the timeout interval is set short enough or the server is busy enough. Saw one of those today.

The warning are there for a reason (although in today's systems with much more memory some are a bit conservatively set) and should not be trifled with.

For the same reason, it is impossible to just give tuning suggestions without knowing exactly what the problem is; this is somewhat akin to going to a doctor and saying "I feel sick" and the doctor asks no further but prescribes Penicillin. Thus just changing UVCONFIG values without knowing the effects of the changes can make a system slow down or, in the worst case, stop functioning.