Since going multilib, boot time really slow
 

Hi,

Just upgraded to multilib as need to be running a specific 32bit program, but since doing so the boot time on my laptop has gone really slow.

Normal boot time to the KDM login screen had previously been around 100 seconds or so. But it's now taking almost 3 x that. Seems to stall around where the fonts are updated, but looking at the system logs, there doesn't seem to be much between 38secs in and 194 seconds.

Any one experienced this and maybe have a pointer in where to start looking for the bottleneck?

TIA

Multilib doesn't affect the boot, unless you made other adjustments. It just adds libs.

Delays are most likely in network log ons. From the snippet of log, this seems to be a big gap

Except this is a similar snippet from the same part of the log from the day previous just before installing the multi lib stuff. As can be seen, the time to the network connection is much less, (132 secs as compared to 219). The only reasonable conclusion is something in them is causing the problem.

Any more ideas?

You may have noticed that if there is a problem starting an application, there is a time delay between trying to start the app, and getting an error message.

Take a look at this:
Start trying to set up the network to getting the error message "link is not ready".

It seems that you need to take a close look at the configuration files to optimize the setting up of the IPv6 network.

Changed nothing in the network settings since forever basically. The laptop is about 3 years old and has been running on the same wireless router since then, through various different releases of Slackware since. I have had no reason to change anything networking wise for that long as it just works.

But, as a question, where do I find the logs for what happens between 38secs and 194 secs. If it helps, that seems to be when runlevel 4 is started, but before X.

I have scanned through every log I can find and they all stop around 38secs and restart around 194 secs, so a big chunk of something going on with no logging info.

Thanks for the suggestions so far.

Usually, the thing that started at 38 seconds (or something else started just before it) continues for a long time. Nothing happened in that period. That's your problem.

Everyone keeps telling me this, but it is simply not yhe case that the network setup is the bottleneck.

While watching the computer boot, the lag seem to be around the mime database update and the part where the gtk-immodules are updated. The PC sits here for ages with the hard drive light flashing indicating some serious HD activity. However, I don't know of a way to log the screen messages to put here.

I have tested the HD with a bootable CD running partition magic as it has some good hard drive tools on it, and it passed OK so it's not a hard drive issue.

So any way to find out what all the HD activity is during the long periods of the thing doing nothing else?

Cheers

You can do this stuff yourself, you know, and test it with time. Go to /etc/rc.d/rc.M around line 200. The logic is simple; if the file is executable, run it. Have you rc.wicd or rc.networkmanager set executable???

You can check the update stuff by just timing the command

Hi there,

Thanks for the tip.

This is what I have when timing the startup scripts from rc.M

That takes 1m28.080s,

Seems to me to be something not quite right there.

I also have another part of the script that take around 10 secs or so,

All the network related stuff, including wicd starts in less than .5 sec.

Any suggestions on speeding this up. I know that plenty of people say not to bother about boot times and ordinarily I would agree with them if I were talking about a server or desktop, but as this is a laptop which gets switched off when not being used it's a bit of a PITA to say the least.

Thanks

That should give 3 times. It might also have to wait for the network to come up. Of course you now have 32 bit modules as well. Just running those from the command line here in a multilib box
Thatś not 1m28s. Go figure.They all are scripts.

Hi again,

the three times are:


These are timings from the last boot up run from rc.M, so a little different from the first set I gave, but you get the idea. I see that your times are really short, so still not quite understanding what's going on.
I'll go and mod the rc.M script to write the results to a file and post it up here to see if you are able to spot anything I might be missing.

Thanks for your patience and assistance so far.

Ok. So go to it. Each of the scripts are like rc.M and you can time whatś going on. There is redundancy, and --verbose is telling you nothing. Youĺl find something, and tell us what it is.

actually I would try to switch off ipv6

Right, as I suspected, it seems to be to do with the gtk-update part of the script as can be seen from the info below:


So in total 47.74 seconds just updating icons?

Does that seem right?

Any ideas anyone as to why this might be and how to correct it?

Thanks

No, it doesn't seem right. I'd google this one
and I'd check that cache, and what's going on. You're nearly there. Maybe it's not writing or reading due to some permissions thing. A cache should be collected and stay. I have multilib, and you have my times