Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

Regards,
Tristan

--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
                   Tristan Miller
Free Software developer, ferret herder, logologist
              https://logological.org/
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

Regards,
Tristan

I think you'll find that the function runs the next time you boot after a kernel update, but I was wondering exactly the same thing yesterday evening.  A new kernel was installed, I rebooted and then watched the system taking a timeout while removing the (-2) kernel.

--
This mail has been tested by https://RKIvirus.com/ and has been found to contain Covid-19. Disinfect after reading.

On 2022-04-14 13:11, Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

AFAIK, it doesn't block here, other things continue running, even the boot sequence. I can not check this instant, but I think I can login while the job is running. I should be able to verify this tomorrow.

You do not say what release you are using.


The job simply calls on zypper to delete the oldest kernel after an update.

You can verify what it does by running:

systemctl cat purge-kernels.service


--
Cheers, Carlos.

Carlos E.R. wrote:

On 2022-04-14 13:11, Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

AFAIK, it doesn't block here, other things continue running, even the boot sequence. I can not check this instant, but I think I can login while the job is running. I should be able to verify this tomorrow.

You do not say what release you are using.


The job simply calls on zypper to delete the oldest kernel after an update.

You can verify what it does by running:

systemctl cat purge-kernels.service

I have Leap 15.3 with the splash screen turned off so that I can see what is going on.  The script runs as part of the boot process and *before* logging in is possible, this is obviously what Tristan sees as well.  In my case - on my old laptop, no ssd - it delays the appearance of the login screen by just over a minute.
No idea if this is something new, I'll often boot and then get on with something else for a minute or two.  It *is* something I first noticed a couple of days before Tristan reported it here.

--
This mail has been tested by https://RKIvirus.com/ and has been found to contain Covid-19. Disinfect after reading.

On 2022-04-17 09:16, Andrew wrote:

Carlos E.R. wrote:

On 2022-04-14 13:11, Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

AFAIK, it doesn't block here, other things continue running, even the boot sequence. I can not check this instant, but I think I can login while the job is running. I should be able to verify this tomorrow.

You do not say what release you are using.


The job simply calls on zypper to delete the oldest kernel after an update.

You can verify what it does by running:

systemctl cat purge-kernels.service

I have Leap 15.3 with the splash screen turned off so that I can see what is going on.  The script runs as part of the boot process and *before* logging in is possible, this is obviously what Tristan sees as well.  In my case - on my old laptop, no ssd - it delays the appearance of the login screen by just over a minute.
No idea if this is something new, I'll often boot and then get on with something else for a minute or two.  It *is* something I first noticed a couple of days before Tristan reported it here.

I booted this Leap 15.3 today after a kernel update. I did:

systemd-analyze plot >bootup.svg
eog bootup.svg


And I clearly see that the boot process continues running while purge-kernels is running (for 41 seconds in my case).

It is the service "display-manager" which waits, apparently.


You can run "systemd-analyze critical-chain", but in my case "purge-kernels" is not listed, which I think it means it does not delay others.


I suggest you ask in the support mail list.


--
Cheers, Carlos.

On Sun, 17 Apr 2022 12:26:14 +0200, Carlos E.R. wrote:

On 2022-04-17 09:16, Andrew wrote:

Carlos E.R. wrote:

On 2022-04-14 13:11, Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute
or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means
uninstalling them.  If this is the case, why is this something that
boot has to block for?  I mean, once the system is up an running, I
can always use zypper or rpm to manually remove old kernels.  So it's
obviously something that *can* be done without interfering with my
use of the machine.  I get why the bootup script might want to clean
up old kernels every once in a while, but why can't it just launch a
process that does this unobtrusively in the background?

AFAIK, it doesn't block here, other things continue running, even the
boot sequence. I can not check this instant, but I think I can login
while the job is running. I should be able to verify this tomorrow.

You do not say what release you are using.


The job simply calls on zypper to delete the oldest kernel after an
update.

You can verify what it does by running:

systemctl cat purge-kernels.service

I have Leap 15.3 with the splash screen turned off so that I can see
what is going on.  The script runs as part of the boot process and
*before* logging in is possible, this is obviously what Tristan sees as
well.  In my case - on my old laptop, no ssd - it delays the appearance
of the login screen by just over a minute.
No idea if this is something new, I'll often boot and then get on with
something else for a minute or two.  It *is* something I first noticed a
couple of days before Tristan reported it here.

I booted this Leap 15.3 today after a kernel update. I did:

systemd-analyze plot >bootup.svg
eog bootup.svg


And I clearly see that the boot process continues running while
purge-kernels is running (for 41 seconds in my case).

It is the service "display-manager" which waits, apparently.


You can run "systemd-analyze critical-chain", but in my case
"purge-kernels" is not listed, which I think it means it does not delay
others.

You can have a large delay if anything runs update-grub.
The large delay is caused by umount of each partition update-grub opened/mounted.

Dear Carlos,

On 17/04/2022 12.26, Carlos E.R. wrote:

I booted this Leap 15.3 today after a kernel update. I did:

systemd-analyze plot >bootup.svg
eog bootup.svg


And I clearly see that the boot process continues running while purge-kernels is running (for 41 seconds in my case).

I'm also running Leap 15.3.  I produced an SVG plot as you suggested and I see that you're right that the boot process continues while the kernels are purged. However, it seems that multi-user.target and graphical.target aren't reached until the purge is complete.  So contrary to what you recalled in your previous message, I can't actually log in until the purge is complete.

Regards,
Tristan

--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
                   Tristan Miller
Free Software developer, ferret herder, logologist
              https://logological.org/
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Tristan Miller wrote:

Dear Carlos,

On 17/04/2022 12.26, Carlos E.R. wrote:

I booted this Leap 15.3 today after a kernel update. I did:

systemd-analyze plot >bootup.svg
eog bootup.svg


And I clearly see that the boot process continues running while purge-kernels is running (for 41 seconds in my case).

I'm also running Leap 15.3.  I produced an SVG plot as you suggested and I see that you're right that the boot process continues while the kernels are purged. However, it seems that multi-user.target and graphical.target aren't reached until the purge is complete.  So contrary to what you recalled in your previous message, I can't actually log in until the purge is complete.

Regards,
Tristan

Carlos also says

It is the service "display-manager" which waits, apparently.

which is - functionally - what we are both saying, the system is not actually useable until the kernels have been purged.  The two of us see the boot-process as being over when we can log in.

--
This mail has been tested by https://RKIvirus.com/ and has been found to contain Covid-19. Disinfect after reading.

On 2022-04-19 14:05, Tristan Miller wrote:

Dear Carlos,

On 17/04/2022 12.26, Carlos E.R. wrote:

I booted this Leap 15.3 today after a kernel update. I did:

systemd-analyze plot >bootup.svg
eog bootup.svg


And I clearly see that the boot process continues running while purge-kernels is running (for 41 seconds in my case).

I'm also running Leap 15.3.  I produced an SVG plot as you suggested and I see that you're right that the boot process continues while the kernels are purged. However, it seems that multi-user.target and graphical.target aren't reached until the purge is complete.  So contrary to what you recalled in your previous message, I can't actually log in until the purge is complete.

Well, I did say in my second post that "display-manager" did not start till "purge-kernels" finished.

So, again, I suggest you bring up the issue in the mail list, because I do not know why "display-manager" waits.

--
Cheers, Carlos.

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

Regards,
Tristan

I have just updated a kernel on my test machine, if you submitted a bug report it has not been acted upon (yet).

--
This mail has been tested by https://RKIvirus.com/ and has been found to contain Covid-19. Disinfect after reading.

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that boot has to block for?  I mean, once the system is up an running, I can always use zypper or rpm to manually remove old kernels.  So it's obviously something that *can* be done without interfering with my use of the machine.  I get why the bootup script might want to clean up old kernels every once in a while, but why can't it just launch a process that does this unobtrusively in the background?

Regards,
Tristan

Having just updated a kernel on a Leap 15.4 machine, the system still waits for the purge to complete.

The purge of the old kernels runs in parallel to the setup of my "wicked managed network interfaces".  On a system with SSD discs the purge takes only slightly longer than the wicked setup.  On my older system the purge takes well over a minute, I'd guess at 80 seconds.

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Andrew <Doug@hyperspace.vogon.gov> wrote:

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute or
two with the message, "A start job is running for Purge old kernels".

If I understand correctly, purging old kernels simply means uninstalling
them.  If this is the case, why is this something that boot has to block
for?  I mean, once the system is up an running, I can always use zypper
or rpm to manually remove old kernels.  So it's obviously something that
*can* be done without interfering with my use of the machine.  I get why
the bootup script might want to clean up old kernels every once in a
while, but why can't it just launch a process that does this
unobtrusively in the background?

Regards,
Tristan

Having just updated a kernel on a Leap 15.4 machine, the system still
waits for the purge to complete.

The purge of the old kernels runs in parallel to the setup of my "wicked
managed network interfaces".  On a system with SSD discs the purge takes
only slightly longer than the wicked setup.  On my older system the
purge takes well over a minute, I'd guess at 80 seconds.

The older system has another problem anyway, my /boot partition is over
500MB and has around 50% free with the current kernel and the -1 kernel.
 This is insufficient when it comes to installing a new kernel and I'm
going to have to start getting rid of the -1 kernel before installing
the new one.  The beast is dual-boot with Windows 10 and I am not
prepared to risk moving the main Windows partion which is just behind /boot.

I also have the impression that the update jobs, e.g. for the locate
database run blocking at that point. Opensuse 15.3.
--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------

On 2022-08-14 22:09, Uwe Bonnes wrote:

Andrew <Doug@hyperspace.vogon.gov> wrote:

I also have the impression that the update jobs, e.g. for the locate
database run blocking at that point. Opensuse 15.3.

Not during boot.

--
Cheers, Carlos.

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #


--
Cheers, Carlos.

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

.... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
.... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default


--
Cheers, Carlos.

Carlos E.R. wrote:

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default

Oh, I did that immediately.  The file was slightly smaller and the partition dropped to 48% used.
Given that the previous 50% was with two kernels, an additional one should still only be 75% before reverting to 50% once the oldest one had been removed, I don't see any alternative to waiting for a new kernel update.
The -1 kernel still has the Plymouth stuff in there so the next update may fail until I remove it, and the following one could then still be ok.  Calculations will not help.

On 2022-08-21 22:11, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default

Oh, I did that immediately.  The file was slightly smaller and the partition dropped to 48% used.

Oh. I expected a significant difference. When I tested this was long ago, though, so I suppose that now they are including so many things in the initrd that it no longer makes an impact.

Given that the previous 50% was with two kernels, an additional one should still only be 75% before reverting to 50% once the oldest one had been removed, I don't see any alternative to waiting for a new kernel update.
The -1 kernel still has the Plymouth stuff in there so the next update may fail until I remove it, and the following one could then still be ok.  Calculations will not help.

Er... no, when you remove the package all the initrds are remade without it. All kernels. Look at the dates of the files (see mine above, same timestamp).

--
Cheers, Carlos.

Carlos E.R. wrote:

On 2022-08-21 22:11, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default

Oh, I did that immediately.  The file was slightly smaller and the partition dropped to 48% used.

Oh. I expected a significant difference. When I tested this was long ago, though, so I suppose that now they are including so many things in the initrd that it no longer makes an impact.

Given that the previous 50% was with two kernels, an additional one should still only be 75% before reverting to 50% once the oldest one had been removed, I don't see any alternative to waiting for a new kernel update.
The -1 kernel still has the Plymouth stuff in there so the next update may fail until I remove it, and the following one could then still be ok.  Calculations will not help.

Er... no, when you remove the package all the initrds are remade without it. All kernels. Look at the dates of the files (see mine above, same timestamp).

I suppose I could have said "you are right" (with the reduced size) but I was waiting for a new kernel to come along.  It has.

240 MB - Total size of /boot
107 MB - Used (with two kernels)
117 MB - Free
This means 48% of /boot is in use.
zypper still said it needed another 10 MB to install the new kernel.
Abort, retry, ignore? [a/r/i] (a):

The adventurous solution would have been to have tried "i" as in "do it anyway".  I copped out, removed the older kernel with Yast -> Software Management (at which point 50-60 MB was in use) and then installed the new one using zypper.  The 240 / 107 / 117 / 48% figures still hold true.
I'm assuming zypper's calculation of how much it needs is at fault, the additional kernel should have meant 160 MB was in use.  zypper's guesstimate of 127 MB for a new kernel seems ridiculous, unless there is some compression going on there during the update process.
There are vmlinux-whatever.gz files in there which take 17-18 MB each, if one takes 80 MB before compression that would explain a lot.

On 2022-09-16 22:52, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-21 22:11, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default

Oh, I did that immediately.  The file was slightly smaller and the partition dropped to 48% used.

Oh. I expected a significant difference. When I tested this was long ago, though, so I suppose that now they are including so many things in the initrd that it no longer makes an impact.

Given that the previous 50% was with two kernels, an additional one should still only be 75% before reverting to 50% once the oldest one had been removed, I don't see any alternative to waiting for a new kernel update.
The -1 kernel still has the Plymouth stuff in there so the next update may fail until I remove it, and the following one could then still be ok.  Calculations will not help.

Er... no, when you remove the package all the initrds are remade without it. All kernels. Look at the dates of the files (see mine above, same timestamp).

I suppose I could have said "you are right" (with the reduced size) but I was waiting for a new kernel to come along.  It has.

240 MB - Total size of /boot
107 MB - Used (with two kernels)
117 MB - Free
This means 48% of /boot is in use.
zypper still said it needed another 10 MB to install the new kernel.
Abort, retry, ignore? [a/r/i] (a):

The adventurous solution would have been to have tried "i" as in "do it anyway".  I copped out, removed the older kernel with Yast -> Software Management (at which point 50-60 MB was in use) and then installed the new one using zypper.  The 240 / 107 / 117 / 48% figures still hold true.
I'm assuming zypper's calculation of how much it needs is at fault, the additional kernel should have meant 160 MB was in use.  zypper's guesstimate of 127 MB for a new kernel seems ridiculous, unless there is some compression going on there during the update process.
There are vmlinux-whatever.gz files in there which take 17-18 MB each, if one takes 80 MB before compression that would explain a lot.

When this happened to me I tried to find 10 meg I could move temporarily to another partition. 240 megs is now too small. Mine is a gigabyte (103M in use), of course after I changed the disk and reformatted.

--
Cheers, Carlos.

Carlos E.R. wrote:

On 2022-09-16 22:52, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-21 22:11, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-20 20:47, Andrew wrote:

Carlos E.R. wrote:

On 2022-08-12 20:02, Andrew wrote:

The older system has another problem anyway, my /boot partition is over 500MB and has around 50% free with the current kernel and the -1 kernel.   This is insufficient when it comes to installing a new kernel and I'm going to have to start getting rid of the -1 kernel before installing the new one.  The beast is dual-boot with Windows 10 and I am not prepared to risk moving the main Windows partion which is just behind /boot.

Try instead uninstalling "plymouth" (and then run mkinitrd).

This package does the graphic display during boot, and it makes the kernel image much larger.

500 M should be enough.

Telcontar:~ # df -h | grep boot
/dev/nvme0n1p4        1011M   98M  862M  11% /boot
/dev/nvme0n1p1         500M   19M  482M   4% /boot/efi
Telcontar:~ #

Thanks, that particular machine is a test system so I tried it.  It still boots with no problems but I won't see if it really helped until the next kernel comes along.
It's my only pre-UEFI system and has been running since 2010, which is why the /boot is sized the way it is.

You can simply do:

df -h /boot

before and after removing the package, to see the effect it has.

This is the file that changes:

cer@Telcontar:~> ls -lh /boot/initrd*

... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.63-default
... 13M Aug  8 14:45 /boot/initrd-5.3.18-150300.59.87-default

Oh, I did that immediately.  The file was slightly smaller and the partition dropped to 48% used.

Oh. I expected a significant difference. When I tested this was long ago, though, so I suppose that now they are including so many things in the initrd that it no longer makes an impact.

Given that the previous 50% was with two kernels, an additional one should still only be 75% before reverting to 50% once the oldest one had been removed, I don't see any alternative to waiting for a new kernel update.
The -1 kernel still has the Plymouth stuff in there so the next update may fail until I remove it, and the following one could then still be ok.  Calculations will not help.

Er... no, when you remove the package all the initrds are remade without it. All kernels. Look at the dates of the files (see mine above, same timestamp).

I suppose I could have said "you are right" (with the reduced size) but I was waiting for a new kernel to come along.  It has.

240 MB - Total size of /boot
107 MB - Used (with two kernels)
117 MB - Free
This means 48% of /boot is in use.
zypper still said it needed another 10 MB to install the new kernel.
Abort, retry, ignore? [a/r/i] (a):

The adventurous solution would have been to have tried "i" as in "do it anyway".  I copped out, removed the older kernel with Yast -> Software Management (at which point 50-60 MB was in use) and then installed the new one using zypper.  The 240 / 107 / 117 / 48% figures still hold true.
I'm assuming zypper's calculation of how much it needs is at fault, the additional kernel should have meant 160 MB was in use.  zypper's guesstimate of 127 MB for a new kernel seems ridiculous, unless there is some compression going on there during the update process.
There are vmlinux-whatever.gz files in there which take 17-18 MB each, if one takes 80 MB before compression that would explain a lot.

When this happened to me I tried to find 10 meg I could move temporarily to another partition. 240 megs is now too small. Mine is a gigabyte (103M in use), of course after I changed the disk and reformatted.

This machine is ancient and is dual boot with Windows 10 (originally Windows 7), I'll have to live with the pain.

On 18.09.2022 at 07:33, Andrew scribbled:

Carlos E.R. wrote:

When this happened to me I tried to find 10 meg I could move
temporarily to another partition. 240 megs is now too small. Mine
is a gigabyte (103M in use), of course after I changed the disk and
reformatted.

This machine is ancient and is dual boot with Windows 10 (originally
Windows 7), I'll have to live with the pain.

Considering that you're also using Windows on that machine, living
with pain should be second nature to you by now. :p

--
With respect,
= Aragorn =

Andrew wrote:

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute
or two with the message, "A start job is running for Purge old
kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that
 boot has to block for?  I mean, once the system is up an running,
I can always use zypper or rpm to manually remove old kernels.  So
it's obviously something that *can* be done without interfering
with my use of the machine.  I get why the bootup script might want
to clean up old kernels every once in a while, but why can't it
just launch a process that does this unobtrusively in the
background?

Regards, Tristan

Having just updated a kernel on a Leap 15.4 machine, the system still
 waits for the purge to complete.

The purge of the old kernels runs in parallel to the setup of my
"wicked managed network interfaces".  On a system with SSD discs the
purge takes only slightly longer than the wicked setup.  On my older
system the purge takes well over a minute, I'd guess at 80 seconds.

The older system has another problem anyway, my /boot partition is
over 500MB and has around 50% free with the current kernel and the -1
kernel. This is insufficient when it comes to installing a new kernel
and I'm going to have to start getting rid of the -1 kernel before
installing the new one.  The beast is dual-boot with Windows 10 and I
am not prepared to risk moving the main Windows partion which is just
behind /boot.

An update has come out which fixes the problem.
purge-kernels-service-0-150200.8.6.1.noarch

It turns out the behaviour was deliberate, YaST -> Software Management -> [search for "purge"], and then look at "Change Log".  The entry for 06 May 2021.

On 2023-03-18 08:35, Don Spam's Reckless Son wrote:

Andrew wrote:

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute
or two with the message, "A start job is running for Purge old
kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that
 boot has to block for?  I mean, once the system is up an running,
I can always use zypper or rpm to manually remove old kernels.  So
it's obviously something that *can* be done without interfering
with my use of the machine.  I get why the bootup script might want
to clean up old kernels every once in a while, but why can't it
just launch a process that does this unobtrusively in the
background?

Regards, Tristan

Having just updated a kernel on a Leap 15.4 machine, the system still
 waits for the purge to complete.

The purge of the old kernels runs in parallel to the setup of my
"wicked managed network interfaces".  On a system with SSD discs the
purge takes only slightly longer than the wicked setup.  On my older
system the purge takes well over a minute, I'd guess at 80 seconds.

The older system has another problem anyway, my /boot partition is
over 500MB and has around 50% free with the current kernel and the -1
kernel. This is insufficient when it comes to installing a new kernel
and I'm going to have to start getting rid of the -1 kernel before
installing the new one.  The beast is dual-boot with Windows 10 and I
am not prepared to risk moving the main Windows partion which is just
behind /boot.

An update has come out which fixes the problem.
purge-kernels-service-0-150200.8.6.1.noarch

It turns out the behaviour was deliberate, YaST -> Software Management -> [search for "purge"], and then look at "Change Log".  The entry for 06 May 2021.

Good find.

Telcontar:~ # rpm -q --changelog purge-kernels-service | less

* Thu May 06 2021 ...@suse.de
- Add ZYPP_LOCK_TIMEOUT=-1 to keep waiting for the lock (boo#1184399).

Which is:

https://bugzilla.opensuse.org/show_bug.cgi?id=1184399

Basically, purge-kernels failed because the user was running zypper at the same time. So they added loop or something waiting for package management to be finished and then start the actual purge.



But this is about a lock condition inside zypper, not about the service holding the boot process. Different issue.




The update you mentions says:


openSUSE-SLE-15.4-2023-793 - Recommended update for purge-kernels-service

This update for purge-kernels-service fixes the following issues:
- Change systemd service type to 'exec' (bsc#1198668)

References:
1198668 (bugzilla) : Tumbleweed: Purge old kernels service block during boot


Comment 11 Michal Koutný 2022-05-13 12:57:12 UTC
....
Thirdly, I can see the purge-kernels.service is of Type=oneshot, which means it'll block subsequent jobs until the ExecStart= command finishes. Not sure if there's actually necessity to wait after kernel removal is done, the Type=simple or Type=exec would only wait for the command start and the removal may run in parallel (even with jobs that specify After=purge-kernels.service).



Curious. We'll see.


--
Cheers, Carlos.

Carlos E.R. wrote:

On 2023-03-18 08:35, Don Spam's Reckless Son wrote:

Andrew wrote:

Tristan Miller wrote:

Greetings.

Occasionally when I boot my machine, the system pauses for a minute
or two with the message, "A start job is running for Purge old
kernels".

If I understand correctly, purging old kernels simply means uninstalling them.  If this is the case, why is this something that
 boot has to block for?  I mean, once the system is up an running,
I can always use zypper or rpm to manually remove old kernels.  So
it's obviously something that *can* be done without interfering
with my use of the machine.  I get why the bootup script might want
to clean up old kernels every once in a while, but why can't it
just launch a process that does this unobtrusively in the
background?

Regards, Tristan

Having just updated a kernel on a Leap 15.4 machine, the system still
 waits for the purge to complete.

The purge of the old kernels runs in parallel to the setup of my
"wicked managed network interfaces".  On a system with SSD discs the
purge takes only slightly longer than the wicked setup.  On my older
system the purge takes well over a minute, I'd guess at 80 seconds.

The older system has another problem anyway, my /boot partition is
over 500MB and has around 50% free with the current kernel and the -1
kernel. This is insufficient when it comes to installing a new kernel
and I'm going to have to start getting rid of the -1 kernel before
installing the new one.  The beast is dual-boot with Windows 10 and I
am not prepared to risk moving the main Windows partion which is just
behind /boot.

An update has come out which fixes the problem.
purge-kernels-service-0-150200.8.6.1.noarch

It turns out the behaviour was deliberate, YaST -> Software Management -> [search for "purge"], and then look at "Change Log".  The entry for 06 May 2021.

Good find.

Telcontar:~ # rpm -q --changelog purge-kernels-service | less

* Thu May 06 2021 ...@suse.de
- Add ZYPP_LOCK_TIMEOUT=-1 to keep waiting for the lock (boo#1184399).

Which is:

https://bugzilla.opensuse.org/show_bug.cgi?id=1184399

Basically, purge-kernels failed because the user was running zypper at the same time. So they added loop or something waiting for package management to be finished and then start the actual purge.



But this is about a lock condition inside zypper, not about the service holding the boot process. Different issue.




The update you mentions says:


openSUSE-SLE-15.4-2023-793 - Recommended update for purge-kernels-service

This update for purge-kernels-service fixes the following issues:
- Change systemd service type to 'exec' (bsc#1198668)

References:
1198668 (bugzilla) : Tumbleweed: Purge old kernels service block during boot


Comment 11 Michal Koutný 2022-05-13 12:57:12 UTC
...
Thirdly, I can see the purge-kernels.service is of Type=oneshot, which means it'll block subsequent jobs until the ExecStart= command finishes. Not sure if there's actually necessity to wait after kernel removal is done, the Type=simple or Type=exec would only wait for the command start and the removal may run in parallel (even with jobs that specify After=purge-kernels.service).



Curious. We'll see.

In the interests of full disclosure I have to point out that the date on the most recent entry in the Change Log (Change service type to exec (boo#1198668)) is 13 May 2022.
Has it really taken 10 months to release that fix?

I was watching out for things on my 10-year-old Laptop (no SSD of course) and saw no sign of the extended wait for a purge which I had seen after previous kernel updates.
The system was running rather slowly after that boot completed, as I expected if a purge was running in the background.
(this was running with the "-1" kernel because the newest one was freezing at a very early stage in the boot process, a different problem).