Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.debian.user > #241165 > unrolled thread

Order of package updates, mysql vs. kernel

Started byVladislav Kurz <vladislav.kurz@webstep.net>
First post2021-10-11 11:40 +0200
Last post2021-10-11 13:20 +0200
Articles 4 — 3 participants

Back to article view | Back to linux.debian.user


Contents

  Order of package updates, mysql vs. kernel Vladislav Kurz <vladislav.kurz@webstep.net> - 2021-10-11 11:40 +0200
    Re: Order of package updates, mysql vs. kernel Josef Strýček <josef.strycek@email.cz> - 2021-10-11 12:20 +0200
      Re: Order of package updates, mysql vs. kernel Vladislav Kurz <vladislav.kurz@webstep.net> - 2021-10-11 13:20 +0200
    Re: Order of package updates, mysql vs. kernel Andrei POPESCU <andreimpopescu@gmail.com> - 2021-10-11 13:20 +0200

#241165 — Order of package updates, mysql vs. kernel

FromVladislav Kurz <vladislav.kurz@webstep.net>
Date2021-10-11 11:40 +0200
SubjectOrder of package updates, mysql vs. kernel
Message-ID<D6ZWN-xp-9@gated-at.bofh.it>
Hello,

during the last point release (and some previous updates too) I have noticed 
one unpleasant issue:

If there is mysql (mariadb) and kernel update at the same time, the order of 
actions is as follows:

stop mysql
compile kernel modules for new kernel (using dkms) 
create new initrd, update-grub, etc...
start mysql

Kernel module compilation may take quite a long time, during which is mysql 
stopped, but does not need to.

I'm not sure whether this should be reported as bug in mysql or kernel 
packaging.

-- 
S pozdravem
        Vladislav Kurz

Centrála: Celní 17/5, 63900 Brno, CZ
Web: http://www.webstep.net
E-Mail: podpora@webstep.net
Tel: 840 840 700, +420 548 214 711
Obchodní podmínky: https://zkrat.to/op

[toc] | [next] | [standalone]


#241166

FromJosef Strýček <josef.strycek@email.cz>
Date2021-10-11 12:20 +0200
Message-ID<D70zv-Ze-1@gated-at.bofh.it>
In reply to#241165

[Multipart message — attachments visible in raw view] — view raw

Dobrý den,



minulý týden se mi stalo něco podobného na testinngu, po aktualizaci, včetně
jádra na 5.14, se mi nechtěl zkompilovat dkms pro nvidii, proto se domnívám,
že za to může nový kernel.




S pozdravem

Josef Strýček




English:






Hello,




something similar happened to testing last week, after the upgrade, 
including the kernel to 5.14, I can't compile dkms for nvidia, so I think 
the new kernel problem.





Good luck

Josef Strýček

[toc] | [prev] | [next] | [standalone]


#241171

FromVladislav Kurz <vladislav.kurz@webstep.net>
Date2021-10-11 13:20 +0200
Message-ID<D71vz-1xl-5@gated-at.bofh.it>
In reply to#241166
Dne pondělí 11. října 2021 12:01:45 CEST, Josef Strýček napsal(a):
> English:
> 
> Hello,
> 
> something similar happened to testing last week, after the upgrade,
> including the kernel to 5.14, I can't compile dkms for nvidia, so I think
> the new kernel problem.

I don't think that it is new problem. I have seen this long time before, but 
just didn't bother to report it.

IMHO it is more about the timing of mysql stop, which comes too early  (at/
before unpack stage). But there may be some good reason for that.

If so then all packages that need to stop their daemon before unpacking the 
new version, should somehow signal apt, that they want to finish the upgrade 
ASAP, before unpacking and configuring other packages. (Instead of the usual 
processing - unpack all, configure all)


-- 
S pozdravem
        Vladislav Kurz

Centrála: Celní 17/5, 63900 Brno, CZ
Web: http://www.webstep.net
E-Mail: podpora@webstep.net
Tel: 840 840 700, +420 548 214 711
Obchodní podmínky: https://zkrat.to/op

[toc] | [prev] | [next] | [standalone]


#241172

FromAndrei POPESCU <andreimpopescu@gmail.com>
Date2021-10-11 13:20 +0200
Message-ID<D71vz-1xl-7@gated-at.bofh.it>
In reply to#241165

[Multipart message — attachments visible in raw view] — view raw

On Lu, 11 oct 21, 10:53:18, Vladislav Kurz wrote:
> Hello,
> 
> during the last point release (and some previous updates too) I have noticed 
> one unpleasant issue:
> 
> If there is mysql (mariadb) and kernel update at the same time, the order of 
> actions is as follows:
> 
> stop mysql
> compile kernel modules for new kernel (using dkms) 
> create new initrd, update-grub, etc...
> start mysql
> 
> Kernel module compilation may take quite a long time, during which is mysql 
> stopped, but does not need to.
> 
> I'm not sure whether this should be reported as bug in mysql or kernel 
> packaging.

Without looking into it I'm guessing MariaDB is stopped in preinst and 
started in postinst, and there might be very good reasons to do it like 
this[1]. It can't however account for whatever else is done by APT/dpkg 
in between[2].

To minimize MariaDB downtime you might want to consider upgrading it in 
a separate APT run from everything else.


[1] README.Debian, the changelog, (archived) bug reports, or comments in 
the package scripts might provide more insight here.

[2] Your issue is triggered by dkms, but it could be any other package 
that does some processing on upgrades, or just the amount of packages 
that you are upgrading at the same time.

Kind regards,
Andrei
-- 
http://wiki.debian.org/FAQsFromDebianUser

[toc] | [prev] | [standalone]


Back to top | Article view | linux.debian.user


csiph-web