Timeline for How do I find what's causing a task to be slow, when CPU, memory, disk and network are not used at 100%?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 28 at 3:48 | comment | added | Basilevs | @GregBurghardt I know. The advice would be different in that situation though. For example - investigating threading issues would be impossible, due to noise of unrelated requests. Profiling and measuring would be very hard in general. I have no advice for production, besides obvious inspection of slow request's execution plan and review of application code, | |
| Oct 27 at 23:30 | comment | added | Greg Burghardt | @Basilevs, often you need to deploy a patch to production with extra logging because you cannot effectively replicate the problem anywhere else. | |
| Oct 27 at 11:21 | comment | added | Basilevs | I did not realize the symptoms are described for production system. Current answers are only applicable to a dedicated test setup. | |
| Oct 25 at 8:09 | comment | added | Basilevs | @ArseniMourzenko network is not used effectively per OP. A DB lock is indeed a potential origin of a bottleneck (hence a suggestion to vary thread count - the perf will be better on lower thread count if lock is a problem). | |
| Oct 25 at 7:41 | comment | added | Arseni Mourzenko | How do I know that it's the inefficient queries that are the problem, and not the network bandwidth or, say, the fact that the database waits for a lock to be released? | |
| Oct 25 at 4:45 | history | edited | Basilevs | CC BY-SA 4.0 | added 11 characters in body |
| Oct 25 at 4:36 | history | edited | Basilevs | CC BY-SA 4.0 | added 280 characters in body |
| Oct 25 at 4:31 | history | answered | Basilevs | CC BY-SA 4.0 |