right, so I think I've had about enough of this...
what causes RPM to (seemingly randomly) lockup when installing or removing packages?
what known fixes are there to stop it from happening anymore?
how does anyone expect any kind of non-advanced technical user to deal with an OS whose central functionality was broken in a major release and not ever fixed, even just before the next major release? grr.
My system: Latest Rawhide (20030921) RPM 4.2.1-0.30 Kernel 2.4.20-19.9
I find it troubling to hear that this is still present, if it's what I think it is.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73097
This started in RH 8, and apparently still exists in RawHide. If this isn't fixed before the next release, that will be 3 major versions shipped with a broken RPM. That is unacceptable.
Sean Middleditch wrote:
right, so I think I've had about enough of this...
what causes RPM to (seemingly randomly) lockup when installing or removing packages?
what known fixes are there to stop it from happening anymore?
how does anyone expect any kind of non-advanced technical user to deal with an OS whose central functionality was broken in a major release and not ever fixed, even just before the next major release? grr.
My system: Latest Rawhide (20030921) RPM 4.2.1-0.30 Kernel 2.4.20-19.9
Justin Georgeson wrote:
I find it troubling to hear that this is still present, if it's what I think it is.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73097
This started in RH 8, and apparently still exists in RawHide. If this isn't fixed before the next release, that will be 3 major versions shipped with a broken RPM. That is unacceptable.
Indeed! Try this:
alias rpm='rm -f /var/lib/rpm/__db.00*; /bin/rpm'
On Mon, 2003-09-22 at 05:24, Paul Gear wrote:
Justin Georgeson wrote:
I find it troubling to hear that this is still present, if it's what I think it is.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73097
This started in RH 8, and apparently still exists in RawHide. If this isn't fixed before the next release, that will be 3 major versions shipped with a broken RPM. That is unacceptable.
Indeed! Try this:
alias rpm='rm -f /var/lib/rpm/__db.00*; /bin/rpm'
Would I be correct in assuming that this won't help in the least for graphical updaters or other apps, as they most likely use librpm? (and wouldn't use that alias anyhow even if they didn't) I almost never use the plain rpm command. :(
On Mon, 22 Sep 2003, Paul Gear wrote:
Justin Georgeson wrote:
I find it troubling to hear that this is still present, if it's what I think it is.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73097
This started in RH 8, and apparently still exists in RawHide. If this isn't fixed before the next release, that will be 3 major versions shipped with a broken RPM. That is unacceptable.
Indeed! Try this:
alias rpm='rm -f /var/lib/rpm/__db.00*; /bin/rpm'
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
Best Regards, Alex.
You may wish to consider posting this to rpm-list@redhat.com , so that Jeff Johnson, RPM maintainer @ red hat, can see / respond / etc to your issue...
- Aaron
rhllist@assursys.co.uk wrote:
On Mon, 22 Sep 2003, Paul Gear wrote:
Justin Georgeson wrote:
I find it troubling to hear that this is still present, if it's what I think it is.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73097
This started in RH 8, and apparently still exists in RawHide. If this isn't fixed before the next release, that will be 3 major versions shipped with a broken RPM. That is unacceptable.
Indeed! Try this:
alias rpm='rm -f /var/lib/rpm/__db.00*; /bin/rpm'
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
Best Regards, Alex.
-- Rhl-list mailing list Rhl-list@redhat.com http://www.redhat.com/mailman/listinfo/rhl-list
On Mon, 2003-09-22 at 10:08, rhllist@assursys.co.uk wrote:
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
The backport you mention are the files in ftp://ftp.rpm.org/pub/rpm/test-4.1.1 ?
Best Regards, Alex.
-- Rhl-list mailing list Rhl-list@redhat.com http://www.redhat.com/mailman/listinfo/rhl-list
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:08, rhllist@assursys.co.uk wrote:
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
The backport you mention are the files in ftp://ftp.rpm.org/pub/rpm/test-4.1.1 ?
Yup, that's the badger. ;-)
Best Regards, Alex.
On Mon, 22 Sep 2003, Aaron Bennett wrote:
You may wish to consider posting this to rpm-list@redhat.com , so that Jeff Johnson, RPM maintainer @ red hat, can see / respond / etc to your issue...
Jeff's well aware of it, and from that Bugzilla report, I understand that he's as frustrated with RH's QA process as the rest of us (in this case, at least)
- Aaron
Best Regards, Alex.
On Mon, Sep 22, 2003 at 09:25:53AM -0400, Sean Middleditch wrote:
Would I be correct in assuming that this won't help in the least for graphical updaters or other apps, as they most likely use librpm? (and wouldn't use that alias anyhow even if they didn't) I almost never use the plain rpm command. :(
How do you know it's rpm locking up then, if you almost never use the plain rpm command? (If up2date tracebacks for some reason and you didn't run it from the command line, it may appear to freeze instead, for instance. I filed that in Bugzilla shortly after the RHL 9 release, although I don't remember the bug number right now.)
Also, could you try running the command "getconf GNU_LIBPTHREAD_VERSION" and tell us the output? (I might have other advice based on the output of this command.)
-Barry K. Nathan barryn@pobox.com
On Mon, 2003-09-22 at 21:43, Barry K. Nathan wrote:
On Mon, Sep 22, 2003 at 09:25:53AM -0400, Sean Middleditch wrote:
Would I be correct in assuming that this won't help in the least for graphical updaters or other apps, as they most likely use librpm? (and wouldn't use that alias anyhow even if they didn't) I almost never use the plain rpm command. :(
How do you know it's rpm locking up then, if you almost never use the plain rpm command? (If up2date tracebacks for some reason and you didn't
Because in the instances I *do* use plain RPM, is *also* locks up. It never locks up when using apt or yum, now that I think on it - just plain rpm and the redhat-install-packages utility. I never use up2date, since it refuses to work w/ Rawhide, only the Severn beta, which isn't what I want.
run it from the command line, it may appear to freeze instead, for instance. I filed that in Bugzilla shortly after the RHL 9 release, although I don't remember the bug number right now.)
I've not found a pattern of crashes and lockups. More often, I'll install a couple packages just fine (usually using redhat-install-packages) and then go to install a another, and get the lockup.
Also, could you try running the command "getconf GNU_LIBPTHREAD_VERSION" and tell us the output? (I might have other advice based on the output of this command.)
elanthis@stargrazer:~$ getconf GNU_LIBPTHREAD_VERSION NPTL 0.57
Running the latest Rawhide. Problem has existed ever since I installed Redhat, version 8.0, after switching from Debian.
-Barry K. Nathan barryn@pobox.com
-- fedora-list mailing list fedora-list@redhat.com http://www.redhat.com/mailman/listinfo/fedora-list
On Mon, 2003-09-22 at 10:46, rhllist@assursys.co.uk wrote:
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:08, rhllist@assursys.co.uk wrote:
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
The backport you mention are the files in ftp://ftp.rpm.org/pub/rpm/test-4.1.1 ?
Yup, that's the badger. ;-)
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists - I'm assuming I probably don't want to downgrade, since the 4.1.1 test is just backported stuff from 4.2, yes?
Best Regards, Alex.
-- Rhl-list mailing list Rhl-list@redhat.com http://www.redhat.com/mailman/listinfo/rhl-list
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:46, rhllist@assursys.co.uk wrote:
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:08, rhllist@assursys.co.uk wrote:
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The backport you mention are the files in ftp://ftp.rpm.org/pub/rpm/test-4.1.1 ?
Yup, that's the badger. ;-)
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists - I'm assuming I probably don't want to downgrade, since the 4.1.1 test is just backported stuff from 4.2, yes?
I have not been able to get it to do this with the rawhide version.. and I have done some pretty darn complicated database transactions with some rpms inside of rpms... Do you have a replicable case including hardware swap size?
On Mon, Sep 22, 2003 at 10:20:36PM -0400, Sean Middleditch wrote:
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists - I'm assuming I probably don't want to downgrade, since the 4.1.1 test is just backported stuff from 4.2, yes?
Yes, it's backported from 4.2-1 (which is newer than the 4.2-0.69 that shipped with Red Hat 9, but older than the 4.2.1 stuff in rawhide).
There's a newer 4.2.1 (I think 4.2.1-3 or so) in the current Red Hat Enterprise Linux update channel right now. Nonetheless, the one in rawhide has also been problem-free for me.
However, I use neither apt nor redhat-install-packages. (I'll probably try redhat-install-packages again after the 2nd Fedora Core beta is released, but not until then.)
I would second the suggestion to mention this on rpm-list. I can't think of anything better to try. (Unless you want to try completely avoiding redhat-install-packages and/or apt, and see if the hangs still happen. Even though you see lockups in rpm itself, one of those might be messing up rpm's locking, and you might not be seeing the effect until well after it happens.)
-Barry K. Nathan barryn@pobox.com
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Mon, 22 Sep 2003 22:20:36 -0400, Sean Middleditch wrote:
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists
Can you describe your usual RPM activities?
Do you interrupt rpm often? Interrupting rpm is one major source of being confronted with stale lock files.
I haven't seen an rpm lockup for a *very* long time. Neither with Shrike nor with Severn. And I have erased, installed and queried hundreds to thousands of packages [in my build environment]. On Red Hat Linux 8.0 it's rpm 4.1.1-1.8x from Fedora "Patches" that works for me:
http://download.fedora.us/patches/redhat/8.0/i386/RPMS.stable/
- --
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 21:43, Barry K. Nathan wrote:
On Mon, Sep 22, 2003 at 09:25:53AM -0400, Sean Middleditch wrote:
Would I be correct in assuming that this won't help in the least for graphical updaters or other apps, as they most likely use librpm? (and wouldn't use that alias anyhow even if they didn't) I almost never use the plain rpm command. :(
How do you know it's rpm locking up then, if you almost never use the plain rpm command? (If up2date tracebacks for some reason and you didn't
Because in the instances I *do* use plain RPM, is *also* locks up. It never locks up when using apt or yum, now that I think on it - just plain rpm and the redhat-install-packages utility. I never use up2date, since it refuses to work w/ Rawhide, only the Severn beta, which isn't what I want.
You can use up2date to get rawhide, as this is basically what the severn update channel supplies.
Also just because rpm won't work it doesn't mean it caused the lockup, the lockup is caused by the previous application not clearing up the locks properly or being killed before it has a chance to do so.
Michael Young
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:46, rhllist@assursys.co.uk wrote:
On Mon, 22 Sep 2003, Sean Middleditch wrote:
On Mon, 2003-09-22 at 10:08, rhllist@assursys.co.uk wrote:
That only removes the problems of lockfiles (__*) being left lying around, it doesn't solve the underlying problems of rpm hanging and subsequent rpm database corruption.
The good news is, that for me at least, the backport of rpm available from http://www.rpm.org works fine. As alluded to in the above Bugzilla report, it's highly annoying that an official erratum hasn't been issued. :-C
The backport you mention are the files in ftp://ftp.rpm.org/pub/rpm/test-4.1.1 ?
Yup, that's the badger. ;-)
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists - I'm assuming I probably don't want to downgrade, since the 4.1.1 test is just backported stuff from 4.2, yes?
Put it this way; I'm running 4.1.1 on RH8 and it's trouble free, despite the heavy use I make of RPM. I haven't tried 4.2.1, and I haven't tried using RH9/Severn as a base platform either. On distros with NPTL-enabled kernels (i.e. RH9+) it may also be well worth setting $LD_ASSUME_KERNEL=2.2.5 (e.g. "LD_ASSUME_KERNEL=2.2.5 rpm -ivh foo-1.2.3-4.i386.rpm").
Best Regards, Alex.
Best Regards, Alex.
Michael Schwendt wrote:
... Can you describe your usual RPM activities?
Do you interrupt rpm often? Interrupting rpm is one major source of being confronted with stale lock files.
Why should that make any difference? If rpm can't handle being interrupted, that means that it is too fragile for the job. Handling user interrupts is one of the most important features rpm can have.
Of course, i'm not saying that this makes the problem any easier to solve, but to require that people never interrupt rpm is to require that they never make mistakes, and we all do that...
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Tue, 23 Sep 2003 21:32:24 +1000, Paul Gear wrote:
Michael Schwendt wrote:
... Can you describe your usual RPM activities?
Do you interrupt rpm often? Interrupting rpm is one major source of being confronted with stale lock files.
Why should that make any difference? If rpm can't handle being interrupted, that means that it is too fragile for the job. Handling user interrupts is one of the most important features rpm can have.
Of course, i'm not saying that this makes the problem any easier to solve, but to require that people never interrupt rpm is to require that they never make mistakes, and we all do that...
Interestingly, nobody requires the user to not interrupt rpm.
[I don't know why I reply to your comment at all. Maybe you don't even want to understand me. ;)]
The difference is that you can avoid frequent lockups when you know what triggers them or what makes them more likely to occur. Similarly, knowledge of the "rm -f /var/lib/rpm/__db* ; rpm -vv --rebuilddb" fix helps most users, too. And the average user is glad about such a work-around once he has managed to run into the rpm lockup trap.
*I*'m not asking anyone to stop complaining. Complain as loud as you can if you feel it's necessary. Look up the corresponding bug reports at bugzilla.redhat.com. Phone up Red Hat customer support and voice your concerns.
- --
On Tue, 2003-09-23 at 04:48, M A Young wrote:
On Mon, 22 Sep 2003, Sean Middleditch wrote:
Because in the instances I *do* use plain RPM, is *also* locks up. It never locks up when using apt or yum, now that I think on it - just plain rpm and the redhat-install-packages utility. I never use up2date, since it refuses to work w/ Rawhide, only the Severn beta, which isn't what I want.
You can use up2date to get rawhide, as this is basically what the severn update channel supplies.
The Severn channel seems to be a lot mroe out of date than the packages I can grab w/ yum. In any case, it's not really that much of a problem to me, I'm not even that fond the up2date UI. ~,^
Also just because rpm won't work it doesn't mean it caused the lockup, the lockup is caused by the previous application not clearing up the locks properly or being killed before it has a chance to do so.
And, again, in many instances that RPM fails, the previous command or app I ran *was* rpm (or redhat-install-packages).
And, even *if* a "previous app" is causing it, it's still happening, which is bull. I don't care what the cause is (from a user standpoint), I only know that it stops working, and the only way to make it work is use "advanced hacker black magic". If there is a propensity for the rpm command or redhat-install-packages to fail (which should be *fixed*), then it should be very gracefully and cleanly handled.
Michael Young
-- fedora-list mailing list fedora-list@redhat.com http://www.redhat.com/mailman/listinfo/fedora-list
On Tue, 2003-09-23 at 03:49, Michael Schwendt wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Mon, 22 Sep 2003 22:20:36 -0400, Sean Middleditch wrote:
Hmm, Rawhide has rpm 4.2.1-0.30, and the bug persists
Can you describe your usual RPM activities?
I run apt or yum once every couple days, but these *never* fail on me, that I can recall. I also download the occasional packages and its dependencies (for example, Abiword 2.0.0), which I generally install by having redhat-install-packages launched from the browser (Epiphany).
I also sometimes run rpm to remove packages, or sometimes install things when redhat-install-packages fails (like it always does when there's dependency issues, since it doesn't seem to bother trying to resolve them from packages I've downloaded in the same folder as the package its trying to install).
Do you interrupt rpm often? Interrupting rpm is one major source of being confronted with stale lock files.
Of course not. I'm not a *complete* goober - were I doing something that stupid, I wouldn't be complaining about rpm sucking. ~,^
The only time I ever stop or kill rpm or redhat-install-packages is when they're locked up.
I haven't seen an rpm lockup for a *very* long time. Neither with Shrike nor with Severn. And I have erased, installed and queried hundreds to thousands of packages [in my build environment]. On Red
Which makes this all rather confusing then. Perhaps there is something in my existing setup that is killing the stability. Guess I'll go the rpm list.
Hat Linux 8.0 it's rpm 4.1.1-1.8x from Fedora "Patches" that works for me:
http://download.fedora.us/patches/redhat/8.0/i386/RPMS.stable/
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.3 (GNU/Linux)
iD8DBQE/b/rv0iMVcrivHFQRAkSpAJ97Ip1NQkhzSkqik84ZWTDGz1otMQCeMma2 9Zgwz4HNUcvEpy6W2VdQirM= =F3o4 -----END PGP SIGNATURE-----
-- fedora-list mailing list fedora-list@redhat.com http://www.redhat.com/mailman/listinfo/fedora-list
On Tue, 2003-09-23 at 08:42, Sean Middleditch wrote:
The Severn channel seems to be a lot mroe out of date than the packages I can grab w/ yum. In any case, it's not really that much of a problem to me, I'm not even that fond the up2date UI. ~,^
Have you visited the RHN website and subscribed your severn system to the redhat-linux-severn-i386-0.-0.93-updates channel in addition to the redhat-linux-severn-i386-0.-0.93 channel?
Gerry
On Tue, 2003-09-23 at 09:56, Gerry Tool wrote:
On Tue, 2003-09-23 at 08:42, Sean Middleditch wrote:
The Severn channel seems to be a lot mroe out of date than the packages I can grab w/ yum. In any case, it's not really that much of a problem to me, I'm not even that fond the up2date UI. ~,^
Have you visited the RHN website and subscribed your severn system to the redhat-linux-severn-i386-0.-0.93-updates channel in addition to the redhat-linux-severn-i386-0.-0.93 channel?
Well, um, no. ^^; I'll give that a try when I get home. Thanks!
Gerry
-- fedora-list mailing list fedora-list@redhat.com http://www.redhat.com/mailman/listinfo/fedora-list
-
Aaron Bennett -
Barry K. Nathan -
Gerry Tool -
Justin Georgeson -
M A Young -
Michael Schwendt -
Paul Gear -
rhllist＠assursys.co.uk -
Sean Middleditch -
Stephen Smoogen