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).

On 2023-03-18 14:53, Don Spam's Reckless Son wrote:
> 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?

If you look at the bugzilla, they released it somewhere in mid 2022 for
Tumbleweed. Then somebody recently mentioned that the bug also affected
15.4, and *then* a patch for 15.4 was also released.

I told here to bring up the issue in the openSUSE mail list. I wrote it
myself (2022-04-20), but the people affected did not speak up and I was
subsequently busy. The next step would be to write a bugzilla.

If you don't write a bugzilla, things are never solved. Things written
here do not exist.

> 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).

--
Cheers, Carlos.

Carlos E.R. wrote:
>
> If you look at the bugzilla, they released it somewhere in mid 2022 for
> Tumbleweed. Then somebody recently mentioned that the bug also affected
> 15.4, and *then* a patch for 15.4 was also released.
>
> I told here to bring up the issue in the openSUSE mail list. I wrote it
> myself (2022-04-20), but the people affected did not speak up and I was
> subsequently busy. The next step would be to write a bugzilla.
>
> If you don't write a bugzilla, things are never solved. Things written
> here do not exist.
>

I tried creating a bugzilla account a year or two ago. Tried. Can't
remember what the problem was, maybe I was not eligible, or maybe there
was a technical problem.

On 2023-03-18 23:31, Don Spam's Reckless Son wrote:
> Carlos E.R. wrote:
>>
>> If you look at the bugzilla, they released it somewhere in mid 2022
>> for Tumbleweed. Then somebody recently mentioned that the bug also
>> affected 15.4, and *then* a patch for 15.4 was also released.
>>
>> I told here to bring up the issue in the openSUSE mail list. I wrote
>> it myself (2022-04-20), but the people affected did not speak up and I
>> was subsequently busy. The next step would be to write a bugzilla.
>>
>> If you don't write a bugzilla, things are never solved. Things written
>> here do not exist.
>>
>
> I tried creating a bugzilla account a year or two ago.  Tried.  Can't
> remember what the problem was, maybe I was not eligible, or maybe there
> was a technical problem.

There is no "eligible" thing.

--
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
>

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 =