Post reply

Slow synchronous SQL Distributed Replay

  • March 10, 2016 at 5:08 pm

    #305621

    I was wondering if anyone has run into the following symptomology:

    I'm using the SQL 2014 distributed replay to test a production workload on new hardware and SQL 2008 R2 SP3. I've set everything up and I decided for my first run I would use the synchronization mode as I figured it should take the same amount of time as the workload I captured (30 minutes.) I kicked off dreplay and it distributed events to the four nodes that I have configured and I'm seeing a reasonable pass rate of >95%. However its not playing it back at the rate I was expecting. Instead it's taking well over 300 minutes!

    I'm using a configuration file from MSDN, and I'm wondering if there is some option that should be tweaked, but I'm at a loss as to which one.

    Update: So I started poking around in the client log file and saw that I've got lots of error messages like this:

    2016-03-10 19:18:51:429 CRITICAL [Client Replay] Active connections exceed 8192, connection 125758 is waiting.

    Which probably points at what's going wrong, but I'm not sure what to do. Connection pooling?

    <?xml version='1.0'?>

    <Options>

    <ReplayOptions>

    <Server>server_name\replay_target_instance</Server>

    <SequencingMode>synchronization</SequencingMode>

    <ConnectTimeScale></ConnectTimeScale>

    <ThinkTimeScale></ThinkTimeScale>

    <HealthmonInterval>60</HealthmonInterval>

    <QueryTimeout>3600</QueryTimeout>

    <ThreadsPerClient></ThreadsPerClient>

    </ReplayOptions>

    <OutputOptions>

    <ResultTrace>

    <RecordRowCount>Yes</RecordRowCount>

    <RecordResultSet>No</RecordResultSet>

    </ResultTrace>

    </OutputOptions>

    </Options>

  • March 10, 2016 at 5:52 pm

    #1864475

    Hmm, finally found this CSS SQL Blog[/url].

  • November 1, 2016 at 8:08 pm

    #1910028

    Hi there.

    I have faced with the same issue some time ago.

    It looks like in a synchronize mode you are unable to replay a large volume of load ('Active connections exceed 8192, ...' was preventing me from being able to continue successfully).

    I was using SQL 2012 SP1 Enterprise on AWS environment trying to replay ~ 10GB of Trace file from production with 12 load boxes.

    Anyway, if you have any idea how to avoid this issue or alter this behavior in order to make it work with large load please let me know.

    Cheers

  • November 2, 2016 at 3:25 pm

    #1910237

    Looking through your commentary there, it looks like you said you have encountered the same warning/error I did.

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply