Problems Updating to Latest Build (affects all but Stable)

[cpanelben]

One of the recent bug fixes altered a widely used function. On some systems, depending upon the Perl version, the fix may result in /scripts/convertemails hanging. This issue is fixed in all affected builds now, but there may be difficulty upgrading your current build. If you encounter this issue please do the following to remedy this issue:

rm -f /scripts/convertemails
/scripts/upcp

A hung upcp run will hang immediately following this point:

root@domain [~]# /scripts/upcp
Found hostname to be XXX.XXXXXX.XXX, which resolves to X.X.X.X
Hostname Virtual Host is already setup
PHP version file is up to date
Ftp Quota Check v1.0
Updating XXXXX:XXXXX....Done
Doing a manual update...
Cpanel updates are coming from layer2.cpanel.net
findutils passes checksum
net-tools passes checksum
Checking /dev/hda....OK
x - creating lock directory
x - SKIPPING /usr/local/lib/php.ini (file already exists)

If your upcp run proceeds past that point, then your system is not affected.

[haze]

Would that essentially be the same as issuing:
/scripts/updatenow;/scripts/upcp or will that somehow not over come this bug ?

[cpanelben]

convertemails is not the cause, but it calls a function in another file that causes the issue. The function can only be updated at a later point in the upcp process. The best course of action is to prevent it from running before the function can be updated by removing /scripts/convertemails.

[jsteel]

How about not being so ambiguous with "On some systems, depending upon the Perl version" and actually outlining what systems and what perl version so we can "head things off at the pass". Letting us know which machines are likely to have the issue will allow us to better prepare when we have *many*.

[cpanelben]

The safest bet is to assume that all your systems are affected and initiate the workaround described above. For such tasks I highly recommend fanout. If you're not familiar with fanout, it's a simple bash script that allows one to execute non-interactive commands to any number of computers at the same time over ssh. It requires that you have ssh keys installed on the computers and ssh-agent running, but it can be a real time saver.

http://www.stearns.org/fanout/

To answer your question more directly, affected systems have a broken implementation of Perl's readlink function. This is a core function, and I was stupified to learn that it could be broken. Here's how it behaves normally:

[user:/]$ ls -l mount1
lrwxrwxrwx  1 root root 9 Apr 29 12:29 mount1 -> /linktest/
[user:/]$ perl -e 'print readlink("/mount1") . "\n";'
/linktest

And Broken:

[user:/]$ ls -l mount1
lrwxrwxrwx  1 root root 9 Apr 29 12:29 mount1 -> /linktest/
[user:/]$ perl -e 'print readlink("/mount1") . "\n";'
linktest

Determining what systems are affected is a little difficult, because sometimes readlink functions properly. So either pre-emptively fixing all your servers or just waiting to see which one's are affected is a difficult decision. At this point it's really a moot point, because if your system hasn't already been affected, then it shouldn't ever be affected.

[philja]

Hello! I have issue with "convertemails" running and eating all CPU resources.

I have removed the "convertemails"
as suggested : rm -f /scripts/convertemails

But I am not sure if I can start : /scripts/upcp , because I have process "convertemails" still running on system. ( I have lowered the priority of "convertemaisls" to lowest so)

[cpanelben]

Running:

killall convertemails

should zap the processes. After running that command, the upcp process may proceed normally. Afterwards, you may want to run upcp once again just to be sure.

[haze]

For what its worth ( if anything ) i've only seen this on one box, and it's running centos 3.4, perl 5.8.4 ( built off layer1 ) .. no others seem to have been affected. This box was also the only centos3.4 cpanel box I have running release ( as opposed to current ).

[easyhoster1]

For what its worth ( if anything ) i've only seen this on one box, and it's running centos 3.4, perl 5.8.4 ( built off layer1 ) .. no others seem to have been affected. This box was also the only centos3.4 cpanel box I have running release ( as opposed to current ).

If anyone is running FREEBSD 4.8-5.3 we have not seen any problems with upcp if you are wondering...

[white sox]

ssh freezes for me at this point

"Hostname Virtual Host is already setup"

my server is very unstable and it seems i can't do anything to fix it.

[theuruguayan]

ssh freezes for me at this point

"Hostname Virtual Host is already setup"

my server is very unstable and it seems i can't do anything to fix it.

hit control+c

it aint a fix, but it will continuo updating..

[jcase]

Actually, i think theres a major memory leak in the latest release. I updated yesterday to release (when will i learn my lesson) from stable, and today there were SEVERAL defunct processes and my swap was at 80% usage. rolled back to stable, and everythings back to normal.

[cPanelBilly]

Not sure what OS you are using becuase you didnt say, however Linux x86 release and stable are exactly the same build.

[So_alegria]

i have a problem! i don't obtain update the cpanel! the upcp stop !
the upcp isn't working! how do i fix this??


root@rsdhost02 [~]# /scripts/upcp
Updating /scripts ...
Sync Source: http://httpupdate.cpanel.net/CURRENT/scripts
Fetching http://httpupdate.cpanel.net/cpanels...cpanelsync.loc
k (0)....@69.90.250.35......connected......receiving...100%......Done
Fetching http://httpupdate.cpanel.net/cpanels...cpanelsync.bz2
(0)....@69.90.250.35......connected......receiving...34%...68%...100%......Done
Done updating /scripts
Found hostname to be rsdhost02.rsdhost.com, which resolves to 69.93.61.210
Hostname Virtual Host is already setup
PHP version file is up to date
Ftp Quota Check v1.0
Updating webgama:rodox....Done
Updating mandalla:bigartes....Done
Doing a manual update...
Cpanel updates are coming from layer2.cpanel.net

[chirpy]

If you're saying that you have the problem that is the topic of this thread, then the fix has been posted above. If not, then you should not be posting to this thread as it doesn't relate to your problem.