Let's start testing out Bullseye + MariaDB 10.4.
We are going to go for Bullseye and MariaDB 10.4 rather than 10.5 or 10.6 in order to minimize the amount of variables that could impact performance.
[x] Compile and package 10.4 for Bullseye
[x] Reimage db1125 replica (test-cluster host) with Bullseye (tested both: keeping `/srv` and wiping it entirely)
[x] Reimage db1124 master (test-cluster) with Bullseye
[x] Reimage db1128 (it was m5 master, but it is now a spare after T288720). Reimage it with bullseye and move it to s1 to let it replicate. Update from 12th Dec 2021: T295965#7580466
[x] Reimage pc2014 (pc1, codfw spare) with Bullseye and let it replicate
[x] Reimage one dbproxy (in codfw) with Bullseye (dbproxy2004 - m5) T295965#7580099
[x] Reimage one codfw sanitarium T295965#7595058
[x] Reimage one codfw multi-instance (db2087 s6 and s7) T295965#7597921
[x] Reimage db2078 (misc multi-instance)
[x] Reimage an external store codfw host (es2032)
[x] Reimage s1 replica and let it serve traffic. (db1128 - originally planned for db1169 but we ran into issues: T299025)
[x] Reimage pc1 master and let it serve traffic (pc1011)
[] Reimage an external store eqiad host (es1022)
[] Possibly, move db1128 to a different section/use it for some other testing (move it to a mX section to make it master at some point)
** currently serving on s1, but to be moved to m1 next week