Lennart Poettering

[Marek Novotny]

Couple interesting reads floating around regarding systemd and Lennart
Poettering.

Google+ post by Lennart Poettering
http://tinyurl.com/prsjl6s

ZD Net (Steven J. Vaughan)
http://www.zdnet.com/lennart-poetterings-linus-torvalds-rant-7000034384/

My own belief of Open Source is that good ideas float and bad ones tend
to sink. I know a few people whom don't like systemd, and truth be told
I am just now starting to learn it. It does boot and shut down almost
instantly. It's pretty fast.

I've read a lot of the concerns regarding systemd and a lot of it comes
down to two things:

1. fear that it will hang a system
2. A belief that it doesn't follow the spirit of UNIX

If it does hang systems then I think it will sink. No one wants unstable
Linux.

I read conflicting posts regarding the second point. Some say much of
what is wanted can be had from systemd as well as init. I don't have a
strong opinion one way or the other as I haven't really dived into
systemd at this point.

comments?

--
Marek Novotny
A member of the Linux Foundation
http://www.linuxfoundation.org
git with the program

[Bit Twister]

On Wed, 08 Oct 2014 11:26:26 -0500, Marek Novotny wrote:

> I am just now starting to learn it. It does boot and shut down almost
> instantly. It's pretty fast.

Just depends on what is done when.

> I've read a lot of the concerns regarding systemd and a lot of it comes
> down to two things:
>
> 1. fear that it will hang a system
> 2. A belief that it doesn't follow the spirit of UNIX

My bitch/complaint is systemd's fault logic testing seems to be pretty
poor at best.

In the past I had my mysql database corrupted on reboot.

Solution, create a script to shut down apps using database, then shut
down mysqld. Now that I have done that, I am seeing a 15 second delay
with named. FRAP, ok add named to apps to shut down. Still seeing
message shutting down.

WTF. Ok, issue a systemctl stop mysqld, then same for named, and run
new_boot_logs.

Well that works so add 18 second delay prior to issuing
the reboot or shutdown mesg.

Double FRAP, sill seeing occasional 15 second timeout on named, and
worse, now seeing 5 minute shutdown of mysqld. You know the one
already shutdown.

Snippet of code from my new_boot_logs script follow:

_cmd=reboot


#****************************
#* shutdown these apps first
#****************************

_servers="mythbackend named postfix mysqld dhcpd httpd"

for _serv in $_servers ; do
_state=$(systemctl is-active $serv.service 2> /dev/null)
if [ $? -eq 0 ] ; then
if [ "$_state" = "active" ] ; then
systemctl stop $_serv
fi
fi
done

(snipped a bunch of code deleting log files and copying /dev/null
into log files)

sleep 18

echo "
Running $_cmd"
sleep 1

$_cmd
exit

#************************ end new_boot_logs ******************************

[Marek Novotny]

On 2014-10-08, Bit Twister <BitTw...@mouse-potato.com> wrote:

[snippy]

> My bitch/complaint is systemd's fault logic testing seems to be pretty
> poor at best.
>
> In the past I had my mysql database corrupted on reboot.
>
> Solution, create a script to shut down apps using database, then shut
> down mysqld. Now that I have done that, I am seeing a 15 second delay
> with named. FRAP, ok add named to apps to shut down. Still seeing
> message shutting down.

[snippy]

Nothing worse then getting corruption. That would freak me out. Is there
a proper channel for you to replay your experiences back to help resolve
this in future updates?

[Bit Twister]

On Wed, 08 Oct 2014 13:48:23 -0500, Marek Novotny wrote:

> Nothing worse then getting database corruption. That would freak me

> out. Is there a proper channel for you to replay your experiences
> back to help resolve this in future updates?

For anything I can reliably recreate the problem, I open a bug report
with the distribution vendor. So far with Mageia
$ grep systemd /local/doc/mdbugs.txt | wc -l
10
have been reported and 8 of those have been resolved.


Currently running systemd-216 on a Mageia Linux Alpha 2 release on two systems.

My test bed running with mythtv does not have the delay problem.
My "Production" system without mythtv had the problem about every 3'rd
or 4'th boot. With latest update it is delaying about every other boot
or two out of three boots.

I have seen intermittent problems like this on systems with lots of
updates. Clean install usually clears them out.

Beta 2 is due out next Tuesday, so about Thursday I should see how it
goes.

I do not see anyone else complaining and I have no desire to
experiment on my "Production MythTV" system to see if I can recreate
the database corruption problem. :)

My problem could be created by the speed of my for loop.
I'll modify it to wait for systemctl stop completion and see if that
clears the delay problem.

[Marek Novotny]

On 2014-10-08, Bit Twister <BitTw...@mouse-potato.com> wrote:

> Beta 2 is due out next Tuesday, so about Thursday I should see how it
> goes.
>
> I do not see anyone else complaining and I have no desire to
> experiment on my "Production MythTV" system to see if I can recreate
> the database corruption problem. :)
>
> My problem could be created by the speed of my for loop.
> I'll modify it to wait for systemctl stop completion and see if that
> clears the delay problem.

I wonder if rhel 7 would do any better for you.

[Bit Twister]

On Wed, 08 Oct 2014 15:16:52 -0500, Marek Novotny wrote:

> I wonder if rhel 7 would do any better for you.

Well, we'll never know. I used to run red hat. :)

Until some manager decided they were losing too much money giving it
away to a bunch of free loaders a long time ago.

Result was they lost a whole bunch of free testers on a whole
lot of different hardware, not to mention a bunch users
running a Usenet help desk for anyone with a problem.

Customer complaints/bugs jumped way up and quality of product dropped.
Usenet help desk became a ghost town.
And as you might have noticed, still not much traffic here. :)

When management looked it to see what happened, they went back to
providing something called fedora I think.

A great majority of those "free loaders" never bothered to come back.

[Marek Novotny]

On 2014-10-08, Bit Twister <BitTw...@mouse-potato.com> wrote:

I just wasn't around for it. I can understand the feeling though. They
really should have thought better about how to deal with all of those
people who were doing right by the community.

I liked Fedora well enough but just like the 6 month release cycle of
Canonical's non LTS I just don't have a desire to keep starting over
every 6 months. Otherwise I'd likely be on it.

[parspes]

I would have to say my main complaint at this time is the amount of
system resources consumed by Systemd, and the fact that is just isn't
neccessary. I appreciate the option but not as a forced "potato on my plate"

Marek Novotny wrote:
> On 2014-10-08, Bit Twister<BitTw...@mouse-potato.com> wrote:
>
> [snippy]
>
>> My bitch/complaint is systemd's fault logic testing seems to be pretty
>> poor at best.
>

> [snippy]
>

[Dan Espen]

I think you have that wrong.
As part of the Red Hat licensing move, they created Fedora.
Same day, same announcement.

> A great majority of those "free loaders" never bothered to come back.

Right after that announcement I started using Fedora.

I have no complaints.

--
Dan Espen