| Summary: | tabled's %check can randomly fail | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Warren Togami <wtogami> |
| Component: | tabled | Assignee: | Jeff Garzik <jgarzik> |
| Status: | CLOSED NEXTRELEASE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
| Severity: | medium | Docs Contact: | |
| Priority: | low | ||
| Version: | 12 | CC: | jgarzik, peterm, zaitcev |
| Target Milestone: | --- | Keywords: | Triaged |
| Target Release: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2010-07-08 04:49:11 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
|
Description
Warren Togami
2009-08-27 15:48:48 UTC
Yes, the tests use a hardcoded TCP port, which may already be in use by something else running on the build box. Unfortunately, we _do_ need to run the tests -- they have definitely caught problems in the past. Suggestions welcome... AF_UNIX maybe? Oh wait, scratch AF_UNIX. It's ouside of our control unless we write the whole new transport for DB4. So, I guess a better option option would be do disable replication entirely and use AF_UNIX between the client and tabled, and tabled and chunkd & CLD. That's a lot of work, because tabled is very high in feedchain. I hoped to avoid AF_UNIX... maybe the tests could figure out a bind port on their own, e.g. 1) bind to * 2) get local address (query port allocated to us by kernel) 3) record this port somewhere, for test clients to use Mucking with netstat is dumb. All we need is cld to write out cld.port alongside cld.pid. The rest of ports can (and should) be fetched from CLD. Both chunkd and tabled register with CLD. And then our own transport would be safe. This means a bunch of code in every client that extracts the service locations from CLD, but we need it in a library anyway. This leaves the replication. I say again, disable it with an option (for both CLD and tabled). This bug appears to have been reported against 'rawhide' during the Fedora 12 development cycle. Changing version to '12'. More information and reason for this action is here: http://fedoraproject.org/wiki/BugZappers/HouseKeeping I think we should be able to close it, now that random port is in. Of course the voting port is still in conflict but we should not see any fallout from that. |