3

We are having an annoying intermittent problem where database connections from our client application on a variety of stations will get dropped. We have applications that use ODBC, Borland Database Engine, and ADO and all get dropped. When connections do get dropped, it isn't everyone - some remain connected while others get dropped. Looking in the logs of one of the applications, I see messages like the following:

  • TCP Provider: The semaphore timeout period has expired
  • TCP Provider: An existing connection was forcibly closed by the remote host
  • Query timeout expired
  • TCP Provider: No such host is known

I don't see any error messages in the SQL error log or in the Windows events on either the client or the SQL Server. Is there anyway to determine if this is a database or network issue from the SQL Server side of things? For example, is there something I can turn on to monitor database connections and get reports when they are dropped? This happens maybe once or twice a day but not at any specific time (not when backups run or something).

The issue seems to be network related and out of scope for this forum.

Improve this question
8
  • Ask your network guys if there are any TCP/IP idle timeouts set on the firewalls.
    – Philᵀᴹ
    Commented Mar 31, 2014 at 13:33
  • @GrumpyPhil: We don't have any firewalls between the clients and the SQL Server other than the Windows Firewall on the client themselves.
    – Caynadian
    Commented Mar 31, 2014 at 13:41
  • @Caynadian and is the behavior any different if you entirely shut that thing off? (In other words, this is certainly sounding like a network connectivity issue, not a database problem.)
    – Aaron Bertrand
    Commented Mar 31, 2014 at 13:46
  • @AaronBertrand: I have it shut off on one particularly problematic machine and it still occurs. In fact, this machine (which is a VM running on VMware 5) experiences this issue more often than most though that might be because it runs 24-7 instead of just business hours. Is there any SQL Server specific network monitoring tools you can think of? I should note that the SQL Server is on a physical host and is not a VM.
    – Caynadian
    Commented Mar 31, 2014 at 13:57
  • It could be a physical problem with the cables. We had a customer with similar complaints about our software. Our IT guy flew to the customer location, replaced some Ethernet cables, and the problem went away.
    – user1008646
    Commented Mar 31, 2014 at 15:09

1 Answer 1

Reset to default
-1

It can be SQL related, not just OS and hardware:

https://support.microsoft.com/en-us/kb/2491214

Improve this answer
2
  • Chris, try to provide actual content and not just link answers.
    – LowlyDBA - John M
    Commented Apr 17, 2015 at 23:59
  • @Chris: Thanks for that but I didn't see any "non-yeilding" error messages in any of the logs. Regardless, it seems to have been a hardware issue as the DB server RAID array crashed and since we have rebuilt the server on a new platform, we haven't had any issues.
    – Caynadian
    Commented Apr 20, 2015 at 16:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.