At the 24 hour run 2015, the publisher responsible for pushing the results over HTTPS with client certificate to the server would sometimes completely lock up without an error message (maybe related to an unreliable connection?) and we had to SIGKILL it and spawn a new one from time to time. We should get to the bottom of this issue.
One thing that was noticed is that the publisher creates a lot of garbage, and there are obvious ways to reduce unnecessary temporary object creation. Maybe this is related to the issue, maybe not.
Other types of publishers (just from database and to file, and receiving updates over HTTPS and writing to a file) did not seem to have any issues.
At the 24 hour run 2015, the publisher responsible for pushing the results over HTTPS with client certificate to the server would sometimes completely lock up without an error message (maybe related to an unreliable connection?) and we had to SIGKILL it and spawn a new one from time to time. We should get to the bottom of this issue.
One thing that was noticed is that the publisher creates a lot of garbage, and there are obvious ways to reduce unnecessary temporary object creation. Maybe this is related to the issue, maybe not.
Other types of publishers (just from database and to file, and receiving updates over HTTPS and writing to a file) did not seem to have any issues.