When using database cache on an Percona XtraDB cluster Truncate's are quite blocking. This means all requests for the cache will wait for the tabel to finish.
When a SST (an initial sync to a new node) is running this can result in waiting for a lock indefinitely which means the sync will fail and the cluster will go down. Other than changing the caching backend to another service, we think not allowing TRUNCATE will aliviate the issues at least. Then we can at least be sure deploying multiple Drupal sites will not result in unfortunate issues where you run into a deadlock situation.
Not too sure what the proper way to fix this would be, but i think removing the possibility for TRUNCATE is at least a start.
| Comment | File | Size | Author |
|---|---|---|---|
| #20 | 3143562-20.patch | 916 bytes | casey |
| #17 | interdiff_3143562_11-17.txt | 188 bytes | ankithashetty |
| #17 | 3143562-17.patch | 1.25 KB | ankithashetty |
| #11 | 3143562-11.patch | 1007 bytes | casey |
| #10 | 3143562-10.patch | 1.4 KB | casey |
Issue fork drupal-3143562
Show commands
Start within a Git clone of the project using the version control instructions.
Or, if you do not have SSH keys set up on git.drupalcode.org:
Comments
Comment #2
bbralaComment #3
casey commentedComment #4
dagomar commentedThis problem also occurs on Drupal 7 sites. Adding a patch.
Comment #6
djdevinAlso applies to AWS Aurora clusters which are probably similar to XtraDB clusters.
Doing things like adding fields would take 60+ seconds. With patch <10s average 5s.
AWS Aurora 2.0 (MySQL 5.7-compatible)
Comment #7
casey commentedComment #10
casey commentedOops, the previous patch was not targeting the mysql driver (for which the driver specific Truncate handler was removed in #3124354: Remove unnecessary classes from the database drivers)
Comment #11
casey commentedComment #15
steinmb commentedLet's see if we still are green
Comment #16
smustgrave commentedCI failures for #11
Comment #17
ankithashettyNot sure if this is the fix to CI errors, but think it'll work...
Comment #18
smustgrave commentedThis issue is being reviewed by the kind folks in Slack, #needs-review-queue-initiative. We are working to keep the size of Needs Review queue [2700+ issues] to around 400 (1 month or less), following Review a patch or merge request as a guide.
This could use an issue summary update using the default issue template. What's the proposed solution? What's the remaining tasks? Example this will need a test case to show the issue
Comment #20
casey commentedReroll for D11.2 and up