Got a box running 4.3 x86 and we're trying up get it updated to 4.6, but yum is consuming all available memory (1GB) and swap (2GB), and then the oom starts killing processes.
I've disabled all the optional repositories, done yum clean all, and still have the same problem even if I try to just update a specific package.
Google got me this link, but no obvious resolution: http://www.centos.org/modules/newbb/viewtopic.php? forum=8&topic_id=5588&viewmode=flat
Any pointers appreciated.
--Chris
On Wed, 2008-06-04 at 16:23 -0500, Chris Boyd wrote:
Got a box running 4.3 x86 and we're trying up get it updated to 4.6, but yum is consuming all available memory (1GB) and swap (2GB), and then the oom starts killing processes.
I've disabled all the optional repositories, done yum clean all, and still have the same problem even if I try to just update a specific package.
Google got me this link, but no obvious resolution: http://www.centos.org/modules/newbb/viewtopic.php? forum=8&topic_id=5588&viewmode=flat
Any pointers appreciated.
Processing a huge number of dependancies (very likely going from 4.3->4.6) will use huge amounts. I would first upgrade yum itself (IIRC, the sqlite changed).
Then, pay attention to the dependancy processing output and pick a few packages at a time to upgrade. After a few passes like this, you should be able to revert to the normal update processing.
Another strategy that might or might not work is to go to runlevel 3 (telinit 3) that eliminates all the graphical stuff and run from a normal VT. If you don't have swap enabled, make a swap partition or file of decent size and try that.
--Chris
<snip sig stuff>
HTH
On Jun 4, 2008, at 4:43 PM, William L. Maltby wrote:
Processing a huge number of dependancies (very likely going from 4.3->4.6) will use huge amounts. I would first upgrade yum itself (IIRC, the sqlite changed).
Tried that with a "yum -y update yum" but we're still getting clobbered.
Then, pay attention to the dependancy processing output and pick a few packages at a time to upgrade. After a few passes like this, you should be able to revert to the normal update processing.
Good suggestion--I'll remember that if I ever get to the point where it's starting to work :-)
Another strategy that might or might not work is to go to runlevel 3 (telinit 3) that eliminates all the graphical stuff and run from a normal VT. If you don't have swap enabled, make a swap partition or file of decent size and try that.
We're already there--no X is installed or running. Swap partition is set to 2GB--I suppose that I could add another swap file and see if that gets me over the hump.
Yum is choking at the point where it print the first "processing primary.xml.gz"
--Chris
On Wed, 2008-06-04 at 16:51 -0500, Chris Boyd wrote:
On Jun 4, 2008, at 4:43 PM, William L. Maltby wrote:
Processing a huge number of dependancies (very likely going from 4.3->4.6) will use huge amounts. I would first upgrade yum itself (IIRC, the sqlite changed).
Tried that with a "yum -y update yum" but we're still getting clobbered.
Did yum or sqlite get updated or did it dump on you there too?
<snip>
Another strategy that might or might not work is to go to runlevel 3 (telinit 3) that eliminates all the graphical stuff and run from a normal VT. If you don't have swap enabled, make a swap partition or file of decent size and try that.
We're already there--no X is installed or running. Swap partition is set to 2GB--I suppose that I could add another swap file and see if that gets me over the hump.
Yum is choking at the point where it print the first "processing primary.xml.gz"
I don't know that much about the yum process. But if it bombs there, I would enable some more verbosity and see if it gives any clues about where in there it is failing. Maybe you have some extra repos enabled that cause excessive memory usage at that step? If that's related (I don't know) you might disable all the repos but the base an updates as a first step and see if at lease the basic stuff can get done.
--Chris
<snip sig stuff>
HTH
On Wed, 2008-06-04 at 16:51 -0500, Chris Boyd wrote:
On Jun 4, 2008, at 4:43 PM, William L. Maltby wrote:
Processing a huge number of dependancies (very likely going from 4.3->4.6) will use huge amounts. I would first upgrade yum itself (IIRC, the sqlite changed).
Tried that with a "yum -y update yum" but we're still getting clobbered.
Then, pay attention to the dependancy processing output and pick a few packages at a time to upgrade. After a few passes like this, you should be able to revert to the normal update processing.
Good suggestion--I'll remember that if I ever get to the point where it's starting to work :-)
Another strategy that might or might not work is to go to runlevel 3 (telinit 3) that eliminates all the graphical stuff and run from a normal VT. If you don't have swap enabled, make a swap partition or file of decent size and try that.
We're already there--no X is installed or running. Swap partition is set to 2GB--I suppose that I could add another swap file and see if that gets me over the hump.
Yum is choking at the point where it print the first "processing primary.xml.gz"
---- kill off any rpm/yum processes... then
rm /var/lib/rpm/__* rpm --rebuilddb
then try yum
Craig
On Jun 4, 2008, at 5:44 PM, Craig White wrote:
kill off any rpm/yum processes... then
rm /var/lib/rpm/__* rpm --rebuilddb
then try yum
Still no joy. Yum appears to be ready to consume all available system memory and swap.
I'm going to try adding a swap file, shutting off the database server and Apache, and then letting it churn.
--Chris
On Wed, Jun 04, 2008 at 07:01:11PM -0500, Chris Boyd enlightened us:
On Jun 4, 2008, at 5:44 PM, Craig White wrote:
kill off any rpm/yum processes... then
rm /var/lib/rpm/__* rpm --rebuilddb
then try yum
Still no joy. Yum appears to be ready to consume all available system memory and swap.
I'm going to try adding a swap file, shutting off the database server and Apache, and then letting it churn.
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Matt
On Jun 4, 2008, at 8:25 PM, Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Good suggestion. Upgraded yum, sqlite, and python. Also added enough swap space via a swapfile to let yum do what it wants. Now it just eats memory until it hits a 4GB wall and then dies.
Version info follows.
[root@hostess ~]# rpm -q yum yum-2.4.3-4.el4.centos [root@hostess ~]# rpm -q sqlite sqlite-3.3.6-2 [root@hostess ~]# rpm -q python python-2.3.4-14.4
--Chris
On Tue, 2008-06-10 at 14:29 -0500, Chris Boyd wrote:
On Jun 4, 2008, at 8:25 PM, Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Good suggestion. Upgraded yum, sqlite, and python. Also added enough swap space via a swapfile to let yum do what it wants. Now it just eats memory until it hits a 4GB wall and then dies.
Version info follows.
[root@hostess ~]# rpm -q yum yum-2.4.3-4.el4.centos [root@hostess ~]# rpm -q sqlite sqlite-3.3.6-2 [root@hostess ~]# rpm -q python python-2.3.4-14.4
Is this a continuation of the massive update seen earlier in this/another thread?
If so, I still think you need to select a *few* of the packages manually and upgrade them to reduce the massive dependancy processing. At some point, enough "foundation" pieces will be in place to run the rest en masse.
It might be instructive to see if swap is actually being used. In another VT, top, free and other utils will report swap usage. Also, swapon with no parameters will report swap usage.
Also, /proc/vmstat might have some useful information (if you dig enough to actually know the meaning of the parameters).
Also, on my cat sys/vm/swappiness has 60 as a value. I know some folks alter this stuff to beneficial/detrimental effect.
--Chris
<snip sig stuff>
HTH
on 6-10-2008 12:29 PM Chris Boyd spake the following:
On Jun 4, 2008, at 8:25 PM, Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Good suggestion. Upgraded yum, sqlite, and python. Also added enough swap space via a swapfile to let yum do what it wants. Now it just eats memory until it hits a 4GB wall and then dies.
Version info follows.
[root@hostess ~]# rpm -q yum yum-2.4.3-4.el4.centos [root@hostess ~]# rpm -q sqlite sqlite-3.3.6-2 [root@hostess ~]# rpm -q python python-2.3.4-14.4
--Chris
There is also a python-sqlite package that might have been updated.
rpm -q yum sqlite python python-sqlite
yum-2.4.3-4.el4.centos.noarch sqlite-3.3.6-2.x86_64 python-2.3.4-14.4.el4_6.1.x86_64 python-sqlite-1.1.7-1.2.1.x86_64
On Jun 10, 2008, at 6:40 PM, Scott Silva wrote:
python-sqlite-1.1.7-1.2.1.x86_64
ding ding ding!
Scott Silva wins a Prize!
That was the last key piece.
William Maltby and Matt Hyclak pointed out the first vital bits of updating yum, sqlite and and then limiting the repos that the system banged on.
Johnny (HappyFunBall) Hughes and Karanbir Singh noted that metadata- parser would need upgrading as well. [Many thanks for the hard work they do on CentOS!]
So, to summarize:
Download the latest RPMS for:
yum yum-metadata-parser sqlite python python-sqlite
You will have to manually satisfy any additional dependencies.
Once done, I was able to "yum -y update" to my heart's content.
--Chris
on 6-10-2008 11:10 PM Chris Boyd spake the following:
On Jun 10, 2008, at 6:40 PM, Scott Silva wrote:
python-sqlite-1.1.7-1.2.1.x86_64
ding ding ding!
Scott Silva wins a Prize!
That was the last key piece.
You know... that was in the thread you linked to in your first message... Last paragraph. You could have been done a week ago.
Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Absolutely. Make sure you get the newer yum with the yum-metadataparser, That helps a lot with ram usage and also speed of the transactions.
Karanbir Singh wrote:
Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Absolutely. Make sure you get the newer yum with the yum-metadataparser, That helps a lot with ram usage and also speed of the transactions.
yum install yum-metadata-parser
Johnny Hughes wrote:
Karanbir Singh wrote:
Matt Hyclak wrote:
If it were me, I would download the latest yum+dependencies and install them via rpm before trying to debug an old yum.
Absolutely. Make sure you get the newer yum with the yum-metadataparser, That helps a lot with ram usage and also speed of the transactions.
yum install yum-metadata-parser
$] rpm -qpR yum-2.4.3-4.el4.centos.noarch.rpm | grep meta yum-metadata-parser
should not need to install it manually, :D the latest released yum in c4 has a direct dep on metadata-parser
Chris Boyd wrote:
Got a box running 4.3 x86 and we're trying up get it updated to 4.6, but yum is consuming all available memory (1GB) and swap (2GB), and then the oom starts killing processes.
I've disabled all the optional repositories, done yum clean all, and still have the same problem even if I try to just update a specific package.
Google got me this link, but no obvious resolution: http://www.centos.org/modules/newbb/viewtopic.php?forum=8&topic_id=5588&...
Any pointers appreciated.
Backup, format, install, restore. ;-)
-
Chris Boyd -
Craig White -
Florin Andrei -
Johnny Hughes -
Karanbir Singh -
Matt Hyclak -
Scott Silva -
William L. Maltby