Bug#979682: startpar: Is it running anything in parallel?

[Michael Krylov]

Package: startpar
Version: 0.61-1
Severity: important

Dear Maintainer,

I get a feeling that startpar doesn't work as it was intended.
That is, it doesn't parallel the service starting at the boot.

I've conducted a couple of experiments: first, with makefile-style boot
and startpar (the default one) and second, without startpar, by creating
the /etc/init.d/.legacy-bootordering file.

Both of them yielded about the same time, 21±1 sec for me.

More than that, after reading its man page, I've tried to run startpar this way:

/lib/startpar/startpar sleep sleep sleep -a 10

And sure enough, it starts three sleep processes one by one and finishes after
30 seconds instead of 10.

I might be wrong, but doesn't this mean that startpar is basically
useless now?

-- System Information:
Debian Release: 10.7
APT prefers stable
APT policy: (500, 'stable')
Architecture: i386 (i686)

Kernel: Linux 4.19.0-13-686 (SMP w/4 CPU cores)
Locale: LANG=C.UTF-8, LC_CTYPE=C.UTF-8 (charmap=UTF-8), LANGUAGE=C.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: sysvinit (via /sbin/init)

Versions of packages startpar depends on:
ii libc6 2.28-10

startpar recommends no packages.

Versions of packages startpar suggests:
ii insserv 1.18.0-2
i

[The Wanderer]

Post by Michael Krylov
Package: startpar
Version: 0.61-1
Severity: important

This version is present in current Debian stable, but there is a newer
version available in current testing and unstable: 0.64-3.

Post by Michael Krylov
Dear Maintainer,

(Disclaimer: I am not a startpar maintainer as such, only an interested
bystander. As such, I have no specific startpar expertise to date, only
what I can observe and deduce.)

Post by Michael Krylov
I get a feeling that startpar doesn't work as it was intended.
That is, it doesn't parallel the service starting at the boot.
I've conducted a couple of experiments: first, with makefile-style boot
and startpar (the default one) and second, without startpar, by creating
the /etc/init.d/.legacy-bootordering file.
Both of them yielded about the same time, 21±1 sec for me.
/lib/startpar/startpar sleep sleep sleep -a 10
And sure enough, it starts three sleep processes one by one and finishes after
30 seconds instead of 10.

On a computer tracking current Debian testing, with startpar 0.64-3, I
see:

$ time /bin/startpar sleep sleep sleep -a 10

real 0m10.006s
user 0m0.009s
sys 0m0.000s

$ time /bin/startpar sleep sleep -a 10

real 0m10.014s
user 0m0.012s
sys 0m0.000s

(Judging from the changelogs, the startpar binary was moved to /bin in
package version 0.63~beta-1.)

As such, I do not appear to encounter this behavior on my system.

Post by Michael Krylov
I might be wrong, but doesn't this mean that startpar is basically
useless now?

* Fixed regression which could cause jobs to run in serial (interactive
mode) instead of parallel when devpts check failed.
Should greatly increase performance on affected systems.

I'd guess that you may be hitting this regression. If so, the problem
should be fixed in the currently latest startpar, although that version
is apparently not in current testing.

[Mikhail Krylov]

Post by The Wanderer
I'd guess that you may be hitting this regression. If so, the problem
should be fixed in the currently latest startpar, although that version
is apparently not in current testing.

Oh, thank you! It was indeed the case. I've installed startpar from
testing and inded, such a command now finished in 10 seconds, and bootup
is like 50% faster.

I guess this ticket can be closed.