[1.7.14] Core Dumps #169
Comments
|
Unfortunately, I didn't have debug symbols being installed, so I just have the callstacks. I think I've installed debug symbols for the program at this point, so if/when it happens again, I may have some more info. |
|
Callstack:
Logs: Footnote: There seem to be a few different PIDs. Not sure if that's expected, or if I'm grabbing a few issues, but only one's creating a core. (?) I'm realizing I don't have Xapian's debug symbols either. Gonna' rebuild that as well with detached debug symbols. I should point out that I've got 11.5 gigs of swap free. And I'm linking against Xapian v1.4.25. |
So it's got a good size document going (40k of terms), but the allocation isn't huge, just 96 bytes. |
|
Weird. |
No, the 96 bytes appears to be just a node in a data structure (red-black tree). So it's a straw-that-broke-the-camel's-back situation. My config has Since I don't have any 1G messages, perhaps there's a memory leak? I'll also try raising But feel free to suggest other things to try to help track down what's going on. Also, I'm not sure if there's a way for you to catch the allocation exception and recover from it. |
|
1G for 40k terms, this seems to much, as each term can not be more than than 245 chars |
|
I found a memory leaj. Can you try latest git ? |
|
Upgraded to latest, and returned to the 1G limit I had before. I'll keep an eye on it for a few days and let you know if I see it again or not. |
|
Not sure if it's related or a new crash, but I just got this twice: Logs: Backtrace: Code for frame #13 + data I can find from the locals: Let me know if you want me to open a different issue for this one. Oh, and if it's relevant, I'm linked against Xapian 1.4.25. |
|
This seems to be either another memory leak, or a low memory on your server. How much memory yo allocated to the indexer ? |
|
Can you provide the few lines before "Can't add document2" ? |
|
please also use latest git |
The additional context you requested before I upgrade to latest. |
|
I meant, can you redo your test with the latest git ? |
|
Any crash ? |
|
I haven't had a crash since August 6th. I've been running the latest since August 13th. (August 13th is also the last time I've restarted my dovecot process.) |
I suspect this one is an upstream bug. |
|
This looks like a memory leak |
|
Could be -- certainly happened when memory was low, but it doesn't look like you're directly causing the crash from what I see. All the pointers you're passing in look like they're non-NULL from what I see. So my gut says that it's upstream not handling an allocation failure properly. I guess, "why not both?" could apply here. |
|
Kindly set verbose=1 and use latest git |
|
Latest has a ton of warnings like this: You should do these like this 👇 if you're just trying to output a code-built string: Otherwise, you could have a With the |
|
Made a PR to fix 'em for you: #171 |
|
Approved your PR. Let's see now the logs before the crash |
Callstack: |
Callstack: |
|
Naturally, there's a ton of log lines. Let me know if there's something I should search for near those core dumps. |
|
Hi @Philippe23 , kindly test latest git |
|
I am seeing the same issue on the 1.7.15 tag weirdly rebuilding on 1.7.14 makes the issue disappear for me. (nixos unstable on commit 915af63a65ff26a44d8897aa90bd9778e56ceb3a, happens with the automated tests of nixos-mailserver commit af7d3bf5daeba3fc28089b015c0dd43f06b176f2 ) |
|
yes, this is a bug in the log debug. I fixed it. Can you try again ? |
|
1.7.16 is fine, thanks. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I'm seeing some core dumps from 1.7.14.
Logs that reference the PID from the core:
A different trace:
Logs:
Another one:
Logs:
I also saw this once in my logs, but didn't find a core for the specific process:
The text was updated successfully, but these errors were encountered: