A
ALX
Hi everybody,
I'm facing a strange problem.
The situation:
- Clients: (about 150) Windows 2000 SP4/XP SP1 with IE6 SP1.
- Servers: two Windows 2000 AS SP4 with NLB and Apache 2.0.47
The servers are used to run a java application (running on WebSphere
Application Server 5.0.2).
The problem is that, from time to time, the browser sends the POST to
the server and immediately after (100-400 ms) a packet with RST flag
enabled, which causes the server to abort the connection.
I sniffed a session between a client and one of the two servers and
the sequence is:
- C->S: SYN
- S->C: SYN+ACK
- C->S: ACK
- C->S: POST ...
- S->C: ACK
- C->S: RST
At this point the client becomes unresponsive, with the hourglass
lasting forever. The only solution is to close the browser and open it
again.
Some notes:
- the servers are not heavily loaded.
- Apache is configured to allow 250 simultaneous connections on each
node.
- most of the clients get the error, so it's not related to the
single client.
- the problem happens in different points of the application, but not
always under the same circumstances and it's not repeatable.
- I've disabled KeepAlive connections on one of the two servers, but
nothing changed.
Has anyone ever seen something like that?
TIA.
ALX
I'm facing a strange problem.
The situation:
- Clients: (about 150) Windows 2000 SP4/XP SP1 with IE6 SP1.
- Servers: two Windows 2000 AS SP4 with NLB and Apache 2.0.47
The servers are used to run a java application (running on WebSphere
Application Server 5.0.2).
The problem is that, from time to time, the browser sends the POST to
the server and immediately after (100-400 ms) a packet with RST flag
enabled, which causes the server to abort the connection.
I sniffed a session between a client and one of the two servers and
the sequence is:
- C->S: SYN
- S->C: SYN+ACK
- C->S: ACK
- C->S: POST ...
- S->C: ACK
- C->S: RST
At this point the client becomes unresponsive, with the hourglass
lasting forever. The only solution is to close the browser and open it
again.
Some notes:
- the servers are not heavily loaded.
- Apache is configured to allow 250 simultaneous connections on each
node.
- most of the clients get the error, so it's not related to the
single client.
- the problem happens in different points of the application, but not
always under the same circumstances and it's not repeatable.
- I've disabled KeepAlive connections on one of the two servers, but
nothing changed.
Has anyone ever seen something like that?
TIA.
ALX