I wrote in a thread entitled "ip_conntrack_ftp fails to load on CentOS4.3" that I had what I considered to be a borked upgrade, using yum update. When I went from 4.2 -> 4.3 I ceased to be able to use the ip_conntrack_ftp module, thus cutting off my users from ftp access.
What I'd like to do is get yum to rerun its update procedure, which should allow any post processing that failed to complete properly to do so. I'm hoping that will add whatever missing piece is causing me the grief.
What I haven't found in my reading is any way to get this to occur. Yum insists that nothing needs to be updated, and the logs seem to show it. Yet here I am with the one server that actually needs to have ftp access, without it. Things were working fine until the fateful yum update run.
I've considered using rpm -F to refresh the files, but from what I can see that won't work. The switch only causes existing packages to be updated. I just don't see any switches in rpm or yum that will enable me to do what I want to attempt. Does anyone know of a wait to cajole yum to do my bidding?
Thanks!
Barry
Barry L. Kline wrote:
I wrote in a thread entitled "ip_conntrack_ftp fails to load on CentOS4.3" that I had what I considered to be a borked upgrade, using yum update. When I went from 4.2 -> 4.3 I ceased to be able to use the ip_conntrack_ftp module, thus cutting off my users from ftp access.
What I'd like to do is get yum to rerun its update procedure, which should allow any post processing that failed to complete properly to do so. I'm hoping that will add whatever missing piece is causing me the grief.
I haven't found anything to do that.
This wont help you now but I ran across this problem a few times over the years with Fedora, RedHat, and the latest time on CentOS installing a fresh copy of 4.2 and then just doing an update a few months ago.
My solution was to create a script that did a 'yum check-update', parse the output, and then automatically do a yum update for each package one by one that was listed. This made sure that if an update failed after installing but before the cleanup stage it would only affect the one package that was being updated(and its dependencies of course). Doing things one by one like this I have never experienced the fail before cleanup problem again though.
The script is not very advanced so it would try to update packages that might have already been updated by dependencies earlier during the script run-time but it just ignores any responses that said a package was already installed(or nothing to do). I had to run it 2 or 3 times for it to upgrade all the packages but eventually there were no more packages to update and I did not experience any problems. This was after repeating the failure after 2 formats.
I wish yum had an option to do this. I would call it --cleanup-after-each-package :).
Any time I do a update that will pull in a bunch of packages(like an upgrade) I use the script.
On Tue, 2006-05-02 at 16:04 -0400, Adam Gibson wrote:
Barry L. Kline wrote:
I wrote in a thread entitled "ip_conntrack_ftp fails to load on CentOS4.3" that I had what I considered to be a borked upgrade, using yum update. When I went from 4.2 -> 4.3 I ceased to be able to use the ip_conntrack_ftp module, thus cutting off my users from ftp access.
What I'd like to do is get yum to rerun its update procedure, which should allow any post processing that failed to complete properly to do so. I'm hoping that will add whatever missing piece is causing me the grief.
I haven't found anything to do that. <snip just in time to be too late to be to be of help! ;-)) >
I'm a rank amateur at this yum/rpm stuff, but maybe ignorant Qs will spark a thought? IIRC, rpm has a status check thingy that will check for missing files, wrong permits, etc. If the yum update really borked and got something into the rpm database as installed completed and that is erroneous, can't you ID the borked components with rpm and then do an install with force of the identified components?
Although I love fully automated everything (NOT!) I never leave myself in a state where only they can do what I want. But I'm a really old CLI guy that has total mistrust of the Graphical Useless Interface.
HTH
William L. Maltby wrote:
I'm a rank amateur at this yum/rpm stuff, but maybe ignorant Qs will spark a thought? IIRC, rpm has a status check thingy that will check for missing files, wrong permits, etc. If the yum update really borked and got something into the rpm database as installed completed and that is erroneous, can't you ID the borked components with rpm and then do an install with force of the identified components?
I have already run that option: rpm -Va. There is nothing in the output that points to my problem. That said, I don't think that rpm -Va would point out any flaws in my installation if the flaws were such that they were the result of the lack of a cleanup script. For example, is a cleanup script creates file X (perhaps a bogus example...) and X was not part of the RPM package list, would rpm -Va be smart enough to note the lack of X? I'm guessing that it wouldn't.
Thanks for the suggestion, though.
Barry
William L. Maltby wrote:
On Tue, 2006-05-02 at 16:04 -0400, Adam Gibson wrote:
Barry L. Kline wrote:
I wrote in a thread entitled "ip_conntrack_ftp fails to load on CentOS4.3" that I had what I considered to be a borked upgrade, using yum update. When I went from 4.2 -> 4.3 I ceased to be able to use the ip_conntrack_ftp module, thus cutting off my users from ftp access.
What I'd like to do is get yum to rerun its update procedure, which should allow any post processing that failed to complete properly to do so. I'm hoping that will add whatever missing piece is causing me the grief.
I haven't found anything to do that. <snip just in time to be too late to be to be of help! ;-)) >
I'm a rank amateur at this yum/rpm stuff, but maybe ignorant Qs will spark a thought? IIRC, rpm has a status check thingy that will check for missing files, wrong permits, etc. If the yum update really borked and got something into the rpm database as installed completed and that is erroneous, can't you ID the borked components with rpm and then do an install with force of the identified components?
Although I love fully automated everything (NOT!) I never leave myself in a state where only they can do what I want. But I'm a really old CLI guy that has total mistrust of the Graphical Useless Interface.
HTH
Missing files are not really the big problem. Extra files leftover from not un-installing the old packages and some number of new packages that did not get the post-install scripts run are the big problems.
Adam Gibson spake the following on 5/3/2006 8:29 AM:
William L. Maltby wrote:
On Tue, 2006-05-02 at 16:04 -0400, Adam Gibson wrote:
Barry L. Kline wrote:
I wrote in a thread entitled "ip_conntrack_ftp fails to load on CentOS4.3" that I had what I considered to be a borked upgrade, using yum update. When I went from 4.2 -> 4.3 I ceased to be able to use the ip_conntrack_ftp module, thus cutting off my users from ftp access.
What I'd like to do is get yum to rerun its update procedure, which should allow any post processing that failed to complete properly to do so. I'm hoping that will add whatever missing piece is causing me the grief.
I haven't found anything to do that. <snip just in time to be too late to be to be of help! ;-)) >
I'm a rank amateur at this yum/rpm stuff, but maybe ignorant Qs will spark a thought? IIRC, rpm has a status check thingy that will check for missing files, wrong permits, etc. If the yum update really borked and got something into the rpm database as installed completed and that is erroneous, can't you ID the borked components with rpm and then do an install with force of the identified components?
Although I love fully automated everything (NOT!) I never leave myself in a state where only they can do what I want. But I'm a really old CLI guy that has total mistrust of the Graphical Useless Interface.
HTH
Missing files are not really the big problem. Extra files leftover from not un-installing the old packages and some number of new packages that did not get the post-install scripts run are the big problems.
Can't you rpm -qa --last and re-install every rpm on the date of the failed yum run?
Scott Silva wrote:
Adam Gibson spake the following on 5/3/2006 8:29 AM:
Can't you rpm -qa --last and re-install every rpm on the date of the failed yum run?
That's a good suggestion. At least it gives me a complete list of everything that was updated, allowing me to go from there. I can then use the --scripts option to RPM to give me something to look at for seeing what is being executed in the cleanup.
Thanks!
Barry L. Kline wrote:
I wish yum had an option to do this. I would call it --cleanup-after-each-package :).
My kingdom for this option!
a bleeding edge version of rpm under the hood will solve this issue for you ( rather, give this feature to you ) - but its going to break pretty much everything on CentOS4's pkg management side of things :)
CentOS 5.... might have this..
Karanbir Singh wrote:
a bleeding edge version of rpm under the hood will solve this issue for you ( rather, give this feature to you ) - but its going to break pretty much everything on CentOS4's pkg management side of things :)
CentOS 5.... might have this..
I think I'll wait for CentOS 5...
-
Adam Gibson -
Barry L. Kline -
Karanbir Singh -
Scott Silva -
William L. Maltby