[MDEV-25793] Update upgrading documentation Created: 2021-05-27 Updated: 2021-09-29 Resolved: 2021-05-27 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Documentation |
| Affects Version/s: | N/A |
| Fix Version/s: | N/A |
| Type: | Bug | Priority: | Major |
| Reporter: | Ian Gilfillan | Assignee: | Ian Gilfillan |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Attachments: |
|
||||||||||||
| Issue Links: |
|
||||||||||||
| Description |
|
See comments in |
| Comments |
| Comment by Otto Kekäläinen [ 2021-06-17 ] |
|
Why is this marked as "Closed"? I didn't spot any place that would have the documentation for DBA's on how to do safe upgrades in the sense as marko suggests they should be done to avoid hitting certain bugs. Until this documentation exists so a DBA can understand it, I cannot understand it either nor complete https://github.com/MariaDB/server/pull/1833. |
| Comment by Otto Kekäläinen [ 2021-06-28 ] |
|
greenman Why is this marked as "Closed"? I didn't spot any place that would have the documentation for DBA's on how to do safe upgrades in the sense as Marko Mäkelä suggests they should be done to avoid hitting certain bugs. To be able to follow-up on bug reports, it would be good that when bugs are closed, there would be some link/trace of how/why it was closed. Thanks! |
| Comment by Ian Gilfillan [ 2021-06-29 ] |
|
I think this was the edit |
| Comment by Otto Kekäläinen [ 2021-06-29 ] |
|
Thanks. Since that link is behind a login-wall, I'll post the update here for clarity: This basically says that run `innodb_fast_shutdown=1` before upgrades to be safe, and the same rule applies for each and every MariaDB version. This seems much simpler than the elaborate description by marko in |
| Comment by Marko Mäkelä [ 2021-06-30 ] |
|
otto, I think that the short version (shut down the old server, using the default value of innodb_fast_shutdown=1) should be good enough for most users. One thing that might be useful to mention is that the DBA should ensure that no XA PREPARE transactions exist in the system before initiating the shutdown. Such transactions are being internally created by the binlog. I am not currently aware of problems related to that. I would expect that no internal XA transactions would never be left behind by the system. Distributed transactions (explicit use of XA statements by the user) were of very limited use until |