January 6, 2003 at 7:21 am
Hello,
I hope anybody can help me because I don't know where to go from know.... The internet doesn't give a solution at all neither do the books online.....
OK we have a server in a replication Scenario.... The links are not that fast but that shouldn't matter at all (Right ?)
I have quesution about SQLPERF(WAITSTATS)
When i Run this query i see the following ....
Wait Type Requests Wait Time Signal
---------------------- --------------- ------------------------ ------------------------
SLEEP 1159.0 1065798.0 1065657.0
RESOURCE_QUEUE 9562.0 4052535.0 1067405.0
NETWORKIO 3834.0 2282560.0 0.0
I know there are network errors because I See them in SP_Monitor..... (But shouldn't TCP take care of these errors ????(what do they Mean))
Furthermore the processor is very busy..... ( sometimes a Processor que length of 27 !!!!!) ( I know add 4 processors) but could the processor be so high because it is managing too many qeued network output ?????
Please help anybody who knows somethingt about SQL.
January 6, 2003 at 8:00 am
Havent tried it, but it should be doable. The log jam will be in the distribution agent posting changes to the subscriber, you may need to decreate the batch and transaction size, increase the timeout. TCP should handle errors in route, but I dont know that it will survive a bad switch port, line cut, etc.
No idea at all why it would drive processor use high. I'd investigate more to see for sure if replication is the culprit and if so, what part? Logreader? Distribution?
Andy
January 6, 2003 at 8:34 am
Andy thanks a Lot for your comments,
I understand most of it But what do you mean with Log Jam ?
And I understood you correct when you say this shouldn't affect server
performance.
Could lowering the 'worker threads'have any impact on this.
I really don't know.
But you agree with me that when i run a Query in the Query analyzer and I get a TDS message TCP should handle that error.
Another question when i run SP_MOnitor i get packet errors... are those TDS errors or Errors from TCP/IP... ( And how High should this be..... is 10 too much or 1 ??? i have 48 in a day is this much or not ?)
And could it be that because the processor is VERY busy sometimes that the network card will suffer from this or are these completely seperate processes...........
January 6, 2003 at 8:41 am
Sorry forgot the last.....
The Distributor and the publisher is the same machine.....
So there are no Agents running
January 6, 2003 at 11:52 am
Sorry, log jam just an expresssion. Dont think worker threads matter. 48 in a day way too many. TDS runs (I think) on top of TCP. I'd be surprised if your card dropped packets due to CPU max out, but possible I'd say. I'll get Steve and/or Brian Kelley in on this.
Andy
January 6, 2003 at 12:14 pm
Not sure of the issue. Could be your NIC requires the CPU (some do, some don't).
What's the replication scenario? Transaction? Merge? Push/pull?
Steve Jones
January 7, 2003 at 2:15 am
It is a mix of SNapshot and Merge replication, it is a PULL replication scenario.
But we have some more info on this...... It seams that the sproc sp_MSAgent_stethoscope is causing the problems..... Not sure of this...
Bu could it be that when something takes longer then 10 minutes to complete the Stethoscope kills the replication. And the stethoscope determine this by examining if there were nmo intries in the log for the last 10 minutes. So a long running qquery ( over 10 minutes) will cause EVERY replication to break ??????
Is this true ?
Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply