-
Notifications
You must be signed in to change notification settings - Fork 23
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. Weβll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Shutting down causes "mremap size mismatch #183
Comments
Maybe related to #81 π€ |
Oh dear me: https://discuss.dgraph.io/t/error-mremap-size-mismatch-on-arm64/15333/7 More worryingly: https://discuss.dgraph.io/t/error-mremap-size-mismatch-on-arm64/15333/8 What kind of data loss does this cause? I'm not sure. And it affects arm* (edit: only 32bit?) arch only due to it not being officially supported? Maybe we can't say |
π¨ π’ Does the version of badger we're using include this recent PR? It seems there have been some arm64-related PR's lately but I'm not yet sure if they address our issues: |
Holy cow, even more π± dgraph-io/badger#1799 So yeh, I think I can pin to the latest commit and see if that helps? I'll get a PR cooking π§βπ³ |
@mycognosist sorry I've got no way of testing this atm... but have got the PRs ready: You'd need to have a local clone of
Then you can run the patched No worries if you don't have time for this atm, it's getting quite involved now π |
Hmmm it looks like they once upon a time had support for older arm / 32 bit but that was on Travis CI which doesn't run anymore... it doesn't seem to be officially supported? The fix in dgraph-io/ristretto#309 was for arm64... so potentially this can resolve the bug on an rpi 4 but not for an orange pi zero? (No pressure to test, just leaving some more breadcrumbs for anyone else diving in...) |
Oh wait a sec, that's a false alarm, they are releasing: https://github.com/dgraph-io/badger/releases/tag/v3.2103.3 So yeah in fact, I think that |
I followed the updated build instructions above (also had to include the dependency replacement line for go-muxrpc). Successful build but crash on execution:
|
@mycognosist Aha, progress! π It seems we need to load in the same Don't have time now but will try tomorrow! Idk if the build tags bubble down into the deps but I will see... i.e. if we pass |
Took a run at it ssbc/margaret#26 You could probably |
I think I followed all of your instructions correctly but unfortunately I'm still seeing the memory errors. On first execution:
On second execution:
Deleting the Steps I followed:
|
@mycognosist Nice! Pity it is still erroring... one thought is - did you start from scratch? I think the logs might be in a weird state now with the mismatched badger config in |
Think it's time for me to dust of a raspberry pi as well and test this thing π§ββοΈ π§ββοΈ π§ββοΈ |
No luck starting from scratch or running repair. I think I should get my RPi4 back online and see if we encounter the same issues on ARM64. Maybe ARM32 is more of a nice-to-have whereas AMD64 and ARM64 are more important targets for support? |
OK, so there is no rpi at hand but I'm running on a debian amd64 machine in my home network and it's been humming away nicely for 2 days without issue. Every time I open up Patchwork on my desktop I see replication running on the https://github.com/ssbc/go-ssb/blob/master/docs/faq.md#what-platforms-does-go-ssb-support does now reference 32 bit dragons also! Hopefully nobody runs into data loss woes while we're still working this out... |
Brilliant! I'll aim to test on my RPi4 this evening / tomorrow so we have some verification of functionality on arm64. |
Ok, testing now. I assume I can simply build |
Ohhhh just got a tasty segfault when I killed the process:
Going to start from scratch and see how I go. |
Yep, can build directly from
Unable to reproduce that segfaulting myself π€ Root of it seems to be:
Doesn't seem to much clarity in what I can dig up on that dgraph-io/badger#522 https://discuss.dgraph.io/t/panic-unclosed-iterator-at-time-of-txn-discard/2855 (?) dgraph-io/badger#1250 (more ?) - unclear to me. |
@mycognosist Still blowing up? Any further luck? π |
I hadn't seen any further blow-ups while testing in the UK. Still need to setup my Pis here in the new spot...need more plug points π |
OK let's close this off for now and if we see it again, can re-open. Thanks for all the testing! |
A new issue to dive into this π€Ώ πββοΈ Would be great to fix for stability reasons...
#124 (comment)
/cc @mycognosist
The text was updated successfully, but these errors were encountered: