I have searched google, dbforums and anywhere else I can think of and am still no further forward with the problem I am having - so if anyone has any advice on this I would be grateful.
The server is a SUN v880 running 3 seperate instances of Ingres 2.6 (installation A runs about 80 users across a 100Mbit network, installation B runs about 40 users across a 100MBit network and installation C runs about 50 users across a 100MBit network).
Installation A and B runs over a couple of routers and firewall as they are offsite to the server and it is both these sets of users that have been experiencing the problems.
All users are affected by disconnections with installation A have the worst of it. There are multiple errors in the errlog.log from the A Ingres server:
IIGCC, 0000002d]: System communication error: Connection timed out.
E_CLFE07_BS_READ_ERR Read from peer process failed; it may have exited.
and these are appearing ramdomly.
The openroad installation on the users machines indicate that the server closed the connection.
We have also monitored the network connections between the server and also the clients and there does not seem to be any noticeable issues. The link does not seem to be saturated in anyway and I can continuous ping across that network to the Ingres server with generally less than 1ms response. However there are some response times larger than 10ms (upto 145ms).
This problem seems to occur randomly and all users did not have a problem at all for several days and it has suddenly started up again.
The only information I have obtained so far is that perhaps Ingres does not like travelling across routers to much and this could be causing the connection failures. Also similiar issues could appear if running across a VPN - which we are not - so suspect it is all pointing to network latency issues.
Does anyone have any suggestions as to how to trouble shoot this?
Let me know if there are any other details you may require.
We had similar issues for a long time on two different versions of Ingres. For us the solution was moving the application to the DB server. We experimented with different network setups and anything we did to reduce latency helped, but it never went away completely.