Indexer Problem: ERROR TcpChannel - Fatal error on accept socket: Software caused connection abort

[chris]

Hi

we get the following error message from time to time on our Indexers (Solaris 10 x86 64bit Splunk 4.2.5 or 4.3):

ERROR TcpChannel - Fatal error on accept socket: Software caused connection abort

If this occurs netstat shows that the port is still listening:


~:netstat -an | grep 99
.9997 *. 0 0 49152 0 LISTEN


But i do not get a connection when telneting to the port, the forwarders can't connect to the indexer anymore and it has to be restarted.

Is there anyone else with the same problem somewhere? Is there a fix for this?

Thanks for helping

Chris

Edit:

We have downgraded to 4.2.4 on one of our indexers, we haven't had the problem since then. But it has only been running for a couple of hours now. I will update this when more time has passed

[chris · Accepted solution]

According to Splunk support this Issue wil be fixed in 4.3.1.
Apparently the root cause of the problem is, that busy indexers can't send TCP Acks fast enough.