Summary
openssl dies after about a day
Metadata
- Id: 4d56733a01322a482dafc73d19fb28b3435bf124
- Trac id: 615
- Type: defect
- Reporter: syn
- Owner:
- Cc:
- Status: closed
- Component: unknown
- Estimated difficulty:
- Resolution: worksforme
- Priority: major
- Milestone:
- Version: 4.7.x
- Changetime: 2012-04-22 14:04:33 UTC
- Created: 2011-06-19 13:51:12 UTC
- Keywords: openssl
Description
Keeping an openssl connection open for about one day leads to the error "SSL socket already closed" being thrown. Since it has no specific condition type it cannot be handled appropriately. The openssl egg should probably not handle this condition specifically anyway but try to write to the closed socket and just let the error raised by the tcp unit propagate.
Changes and comments
[2011-06-19 13:52:23 UTC] sjamaan wrote:
I think the main problem is that it's reinvoking startup after the socket & tcp connection has been marked closed. It shouldn't do that.
[2011-07-17 17:13:05 UTC] sjamaan wrote:
The only way I'm able to reproduce this is if I catch and ignore the handshake timeout exception and then re-use that same ssl port.
I think this may be a bug in your code. Could you post a (possibly simplified) version of it that fails in this way?
[2011-07-17 17:28:21 UTC] sjamaan wrote:
Also, you may be accidentally reading from or writing to an already closed port. Try the patch in #653 to see if that's happening here.
[2011-10-06 09:53:36 UTC] felix removed milestone 4.8.0
[2012-02-15 22:28:56 UTC] sjamaan wrote:
Is this still an issue with the refactorings Thomas made with 1.6.1?
[2012-02-15 23:45:13 UTC] mario wrote:
It's been working for me. I have a server that has been up for weeks now. I'm using openssl 1.5.1, though.
[2012-04-22 14:04:33 UTC] syn changed status from new to closed
[2012-04-22 14:04:33 UTC] syn set resolution to worksforme
[2012-04-22 14:04:33 UTC] syn wrote:
Unfortunately I'm unable to reproduce this anymore as I don't have the code at hand. Since Mario says it works for him I'm closing this issue for now.