| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Use find_recordset when looking up recordsets internally https://review.opendev.org/c/openstack/designate/+/879305 | 05:55 |
|---|---|---|
| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Secondary zone loops AXFR transfer during zone creation https://review.opendev.org/c/openstack/designate/+/864131 | 05:55 |
| ozzzo_work | eandersson: Switching from get_recordset to find_recordset didn't make a difference, so I tried deleting thisrecord instead of record_to_delete, and that fixed it: https://paste.openstack.org/show/bbKOfD6JNWHf9fD4qlcW/ | 12:08 |
| ozzzo_work | Apparently there is a difference between the recordset returned by find/get, and the recordset that is returned by stepping through the record | 12:09 |
| ozzzo_work | It's not obviouis what the difference is; if I use json.dumps to print both, they appear identical | 12:09 |
| ozzzo_work | sorry that was kind of garbled. When _delete calls _update_or_delete_recordset, it passes "record" | 12:19 |
| ozzzo_work | It appears that the "record" that is passed by _delete (setting record_to_delete) is different from the "thisrecord" that is generated by stepping through recordset.records | 12:20 |
| ozzzo_work | but only in rare cases | 12:21 |
| ozzzo_work | approximately 1 out of 50 | 12:22 |
| eandersson | Interesting - I did extensive testing on current version, and I’ll retreat that on train to see if it is Train specific | 15:44 |
| eandersson | *retry | 15:45 |
| opendevreview | Michael Johnson proposed openstack/designate master: Multi-pools implementation https://review.opendev.org/c/openstack/designate/+/857978 | 15:58 |
| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Use ids to remove records in the sink https://review.opendev.org/c/openstack/designate/+/879832 | 22:50 |
| eandersson | ozzzo_work ^ try the above, I assume this is the same as you are talkinga about right? | 22:51 |
| eandersson | So a record in a different state will fail | 23:12 |
| eandersson | Can you create a bug for this? | 23:12 |
| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Use ids to remove records in the sink https://review.opendev.org/c/openstack/designate/+/879832 | 23:15 |
| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Use ids to remove records for sink and ptr https://review.opendev.org/c/openstack/designate/+/879832 | 23:19 |
| eandersson | Great find btw. I was convinced it was the find operations that weren't returning the records, but it's just the status that changes during the process and removal does not support that. Much like johnsom suspected at some point. | 23:27 |
| eandersson | It's a very small window for the race condition, but enough I guess for your setup to hit it every ~50th time. | 23:27 |
| eandersson | https://paste.opendev.org/show/bY9rEhkdTl1Soo6GD0NQ/ | 23:29 |
| eandersson | Something like this would reproduce it | 23:29 |
| opendevreview | Erik Olof Gunnar Andersson proposed openstack/designate master: Use ids to remove records for sink and ptr https://review.opendev.org/c/openstack/designate/+/879832 | 23:45 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!