-
Notifications
You must be signed in to change notification settings - Fork 1.2k
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
Scylla core dump on EC2, reporting Checksum error #598
Comments
isn't this a dup of #593 @tzach please note that you are replaying commitlogs - so it mean the On Mon, Nov 23, 2015 at 4:41 PM, Tzach Livyatan notifications@github.com
|
Feel free to close this issue if its duplicate.
Not on purpose
I'm guessing the second is more likely. |
On Mon, Nov 23, 2015 at 4:54 PM, Tzach Livyatan notifications@github.com
In this case its strange that this happend - as the only items that should
If there is no additional core then its likely not the case.
|
Look like the core dump is the original issue, and the commit log check sum errors happend after restart (again and again). |
I reproduce the problem, (or a problem) by
the core new dump
|
This looks like a live process, all the threads are where you expect Or was the list truncated? There should be 2 threads per lcore. On 11/23/2015 06:35 PM, Tzach Livyatan wrote:
|
A user report the same issue with 0.13 AMI |
You are running Scylla AMI 0.12. It apparently does not have the changes to make CRC failures non-fatal. Whether or not the commit log segments should be corrupted or not is another issue, but assuming a harsh kill, and the above pointing to actual data sections having been incompletely written, it is not that strange really. |
But having said that, and gathered feet in my mouth, I do see a pretty obvious bug in the replay iterator that might have a little something to do with the issue (false crc errors). |
Customer error, with similar logs is from AMI 0.13 |
So, good news is that with a fix for a file position issue, I can, as far as I can tell, read the log segments fine. I'll send the patch for the commit log reader issue, but I think the marshalling might indicate some other compatibility issue. |
Are you reading the commitlogs,sstables created by 0.13 using head (with your patch) ? If so can we try and do a test using 0.13 and your patches - does that still have an issue ? |
Cherry-picking the commit log fix onto 0.13 lets me start scylla with the data dump and no errors/printouts. |
yes - based on Calle's test - untill we get a different sample that causes an issue |
Using Scylla AMI 0.12
A stress run result in scylla service goes down
Last messages on journal
core sump
The text was updated successfully, but these errors were encountered: