Skip to content
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

[Bug](backup) publish will always timeout after doris upgraded from 1.2 to 2.0 #37076

Open
3 tasks done
xy720 opened this issue Jul 1, 2024 · 0 comments
Open
3 tasks done

Comments

@xy720
Copy link
Member

xy720 commented Jul 1, 2024

Search before asking

  • I had searched in the issues and found no similar issues.

Version

1.2|2.0

What's Wrong?

After upgrading from version 1.2 to version 2.0, we found that all tables writes failed during the publish phase.

E.g. Do insert on a old table upgraded from 1.2, it will encountered publish timeout(default 10s):

MySQL [report]> INSERT INTO t_jishu_silian (id, jflg, compid, empid, jishu, bilv1, bilv2, dayin, dayout) VALUES('598690564658692098', 1, '585500694313795584', NULL, NULL, 0.60000, 0.00000, NULL, NULL);
Query OK, 1 row affected (10.03 sec)
{'label':'insert_1fd3d610d1ec489b_b0621f9b4ae7aac5', 'status':'COMMITTED', 'txnId':'2009'}

After reviewing, we found that as long as backup and restore operations were performed on the original 1.2 version cluster, it would lead to publish failures after upgrading to version 2.0.

The reason is that during the restore operation in version 1.2, when using snapshot to replace the old data file, the tablet metadata used the information in the backup file, then it is modified according to the real meta in FE, but the table/partition ID of the tablet metadata was not modified.

Therefore, after the restore operation, the table/partition ID in the tablet meta is from the backup file, rather than the real meta in FE.

After upgrade to 2.0, the table/partition ID between tablet meta in BE and real meta in FE is mismatch, which will cause the publish failures.

What You Expected?

To completely fix this issue, we need to take a few steps:

  1. Fix the backup and restore missing modifying table/partition ID of version 1.2.
  2. Add a feature of checking and correcting table/partition ID in 1.2/2.0, to solve the problem that table/partition ID mismatch after upgrading.

How to Reproduce?

1、Preform a restore operation in a 1.2 version cluster.

2、upgrade cluster from version 1.2 to 2.0.

3、Do insert on a table which has been done a restore operation in step 1.

4、publish timeout

Anything Else?

No response

Are you willing to submit PR?

  • Yes I am willing to submit a PR!

Code of Conduct

@xy720 xy720 changed the title [Bug] publish will always fail after doris upgraded from 1.2 to 2.0 [Bug](bakcup) publish will always fail after doris upgraded from 1.2 to 2.0 Jul 1, 2024
@xy720 xy720 changed the title [Bug](bakcup) publish will always fail after doris upgraded from 1.2 to 2.0 [Bug](backup) publish will always fail after doris upgraded from 1.2 to 2.0 Jul 1, 2024
@xy720 xy720 changed the title [Bug](backup) publish will always fail after doris upgraded from 1.2 to 2.0 [Bug](backup) publish will always timeout after doris upgraded from 1.2 to 2.0 Jul 1, 2024
xy720 added a commit that referenced this issue Jul 15, 2024
## Proposed changes

Issue Number: step 1 in version master in #37076 

<!--Describe your changes.-->
xy720 added a commit that referenced this issue Jul 15, 2024
## Proposed changes

Issue Number: step 1 in version 2.0 in #37076

<!--Describe your changes.-->
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant