I have JIRA 5.0.6 running with GreenHopper and Bonfire. I'm wrestling with an issue where HTTP responses for most write operations on most projects take about 20 seconds.
I know that it's the response that's hanging because I can immediately refresh the same issue in another window/tab and the issue is updated, whereas the original window/tab is still waiting for an HTTP response from the server.
Some projects aren't susceptible at all to this issue. And for the projects that are, adding people to an issue's watchlist is instantaneous.
I've already ruled out the database as a bottleneck; I initally thought the InnoDB buffer pool size was too small, but increasing it did nothing. The databse has since been migrated to a dedicated database server on a different machine, and it's using MS SQL instead of MySQL now. Connection pools are plentiful; nothing running low on resources. No performance improvement.
I've already ruled out memory as a bottleneck; I've increased max values on everything: heap size, PermGen space, etc. JVM memory graphs look fine.
I've already ruled out antivirus as a bottleneck; I completely disabled all antivirus scanning and nothing changed.
I enabled profiling briefly to determine what was happening. The log didn't show anything useful. Instead, there was just a substantial gap between one log entry and the next. (It's been a while, so I don't remember specifics anymore.)
Anybody else experience anything like this, or maybe have any ideas about what's going on here? Any chance this could be something event-related? (And if so, why aren't events handled asynchronously?)