The tidying up of loose ends
Nov. 6th, 2013 08:24 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
sawylTaking advantage of an unexpected lull in the workload, I made a last minute decision to take a few days off. Consequently, I've had an extremely productive day packaging up and writing up things for the others to look after in my absence; which, I noticed while I was doing it, exactly mirrors my actions of a year ago when I had to package up a similar test case for a similar problem for the others to look after while I was away!
I was even able to come up with what I hope will turn out to be a breakthrough in a particularly tricky memory leak problem I've been working on for much of the last year. Concentrating on the fact that the problem occurs frequently with one specific type of workload and rarely with the others, I did what I should have done in the first place and focused on what differentiated the two. After repeating a couple of previous tests to rule out the obvious, I noticed that we were using a command with a specific set of options on the system where the problem was seen and when I repeatedly ran the same command on the test machine, I was able to get one of the daemons to increase its memory footprint by a couple of megabytes.
While I'm not entirely convinced that this is the same as the problem we've been struggling with — correlation does not equal causation! — I'm pleased that I've been able to demonstrate a memory leak in the code and I'm hopeful that this might at least point the support people in the right direction.
I was even able to come up with what I hope will turn out to be a breakthrough in a particularly tricky memory leak problem I've been working on for much of the last year. Concentrating on the fact that the problem occurs frequently with one specific type of workload and rarely with the others, I did what I should have done in the first place and focused on what differentiated the two. After repeating a couple of previous tests to rule out the obvious, I noticed that we were using a command with a specific set of options on the system where the problem was seen and when I repeatedly ran the same command on the test machine, I was able to get one of the daemons to increase its memory footprint by a couple of megabytes.
While I'm not entirely convinced that this is the same as the problem we've been struggling with — correlation does not equal causation! — I'm pleased that I've been able to demonstrate a memory leak in the code and I'm hopeful that this might at least point the support people in the right direction.
