-
Notifications
You must be signed in to change notification settings - Fork 24
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
1.5.3 crashing repeatedly starting about a hour ago #65
Comments
Ok. Quick question: Do you know if this also happens with 1.5.2? |
Sorry, cant check it. |
Same on another box, with 1.4.2:
|
Thank you. |
I have 3 boxes with fort. |
Not sure if there is a relation, but @job tweeted the following earlier today:
|
Unfortunately I can not get fort to finish the initial sync when firewalling the hostname of the repo containing the P256/bgpsec certificate. This does however prevent the crash. Next step would be to serve another rsync repo at that IP and see what happens. |
Is the stack trace really the only output? |
Yes, it is. No error message. |
Started it with debug argument, last lines from syslog: Nov 10 01:00:28 nuff fort: DBG: Client closed the socket. |
That stack trace looks very different, though. (It's not a fatal one.) |
1.5.0:
1.5.3 passes on my machine (now). Of course I do not know if the same content was served to this version: A repository operator could serve different content depending on the user-agent etc. Curiously, valid router keys is 0 while there is one in the repo:
|
Ok. We do have the error message, though. Thanks. |
Normal. It's supposed to skip BGP certificates, since issue58-proper hasn't been merged. |
If I run
In the cache dir after running 1.5.3 I do not see the P256/bgpsec certificate. With 1.5.0 I do. My routinator also sees two valid bgpsec certificates. It's late in my local timezone so I will not dive deeper - but I would expect the files to still be in the repo? Could be fort cleaning the directory. Could be a split view being served depending on the user-agent. |
Does anybody happen to still have a copy of the cache from when the crash happened? |
Local repository is in https://drive.google.com/file/d/13SVrMK4xDENcS6kqBI3wNtUo0yOD29z0/view?usp=sharing |
One thing I failed to mention in the release notes is that, since 1.5.3, Fort is now more trigger-happy with the stack traces. I thought this might help since bug reports have been on the rise. But it's too trigger happy, evidently. You can tell simply by running the binary without arguments:
I'm going to have this toned down by the next release. Just to be clear: Bug #65 refers to this critical:
|
As stated in the 1.6.0 release notes, I have found and patched several instances of undefined behavior during the reviews. There is no way to prove that these caused this particular crash (particularly considering that I never managed to reproduce it, and the OP already probably left), but the code has changed so much, at this point I expect the bug to manifest in a completely different way, if at all. Please upgrade to the latest version. If it crashes again, please open a new bug. |
fort 1.5.3 (installed from fort-1.5.3-1.el8.x86_64.rpm)
Started to crash repeatedly beginning about a hour ago.
Unable to get it working.
Tried to wipe /var/fort_cache, did not help
The text was updated successfully, but these errors were encountered: