Red Hat Bugzilla – Bug 77467
upgrade to RH 8.0 glibc breaks mysql remote connections
Last modified: 2016-11-24 10:09:01 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20021003
Description of problem:
Remot connections to mysql fail. This is exactly the same as the folowing bug
and errata for pre 8.0. Why no fix for 8.0?
Bug # 75128
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.Try a remote connection to an 8.0 box running mysql.
Actual Results: Connection fails, just like in the bug for other versions of
Expected Results: Remote connections should work.
I've been watching this bug for some time waiting for a solution, and when the
solution finaly came, it was for all of the older versions of redhat. I guess
I missed the fine print (version numbers) in the other bug reports. Probably
others did as well. Any chance this can get fixed soon as well?
You might want to upgrade to MySQL-3.23.53a-1. :)
For 8.0, you can use rpms from rawhide for the time being.
It will eventually become errata when one further regex bugs gets fixed.
*** Bug 77039 has been marked as a duplicate of this bug. ***
*** Bug 78990 has been marked as a duplicate of this bug. ***
*** Bug 74943 has been marked as a duplicate of this bug. ***
*** Bug 75290 has been marked as a duplicate of this bug. ***
*** Bug 78769 has been marked as a duplicate of this bug. ***
*** Bug 76229 has been marked as a duplicate of this bug. ***
creating symlinks from /usr/lib/mysql/libmysqlclient.so so.10 so.10.0.0 to
/usr/lib and recompiling the RPM with the release 18 kernel and latest glibc
updates. seams to have resolved the issue with connecting to the mysql port from
remote without having entrys in /etc/hosts this also resolved the annoyance
with compiling 3rd party apps that were looking for the shared lib in /usr/lib
even though /usr/lib/mysql is in /etc/ld.so.conf and ldconfig being exicuted.
Although new rpm for RH 8.0 is just released, it still doesn't address the
remote connect problem!?
I'm hoping I'm not spamming a list or posting in an inappropriate spot but:
On RH8.0, MySQL 3.23.49, glib-1.2.10-5 I was able to fix the remote connection
problem by adding this line to /etc/my.cnf in the [mysqld] section:
set-variable = thread_stack=256k
this suggestion was placed elsewhere in the string of bug reports related to
the probem but in relation to previous versions of RH.
Hope this helps.
The above didn't work for me, but adding
to the [mysqld] section in /etc/my.cnf did (as mentioned in bug 75128 comment 40
relating to earlier RH versions). Unfortunately, it means changing the hosts in
the user table, but at least it now works.
Can duplicate as well.
Server: RH8, MySQL compiled from source (source from mysql.com not SRPM)
Client: RH8, MySQL RPMS for client
DNS running on server.mydomain.org. I can do forward and reverse lookups on
client.mydomain.org. MySQL table permissions should allow for connections from
client.mydomain.org (host specified, not IP).
I just installed RH 8.0 and MySql which comes with it (3.23.52). I used the
MySqlCC GUI (downloaded from mysql.com) to connect to MySql on the local host,
no problem. I tried to connect from another machine (Windows), with the error
2013 Lost Connection...
I first confirmed security settings, which were OK (no firewall).
For some reason, the only way the remote connection would work was if I entered
the client workstation's info in the HOSTS section (System Settings > Network >
Hosts)of the RH8/MySQL server. The DNS which the server uses is fine (all
other apps requiring DNS lookup work fine; forward and reverse tables all OK),
but for some reason MySql does not like it. Unfortunately, you would have to
make HOSTS entries for all the workstations from which you want to connect to
the MySql server. Actually, I only have a couple workstations anyway.
Being new (~ couple days) to Linux/MySQL, I also did not realize that the user
account HOST field was for the CLIENT host name, not the server host name. I
thought there was a correlation between the user account HOST field and the
HOST field in the MySqlCC connection registration (in the MySqlCC connection
registration, HOST is the host name of the MySQL server to which you want to
connect). Oh well, live and learn. The user account HOST name can be a
wildcard so you can connect from any workstation, but I kept unknowingly using
the server host name for HOST when I created the user accounts. This gave me a
1130 error until I figured it all out.
I hope this helps someone.
Anyone from redhat is going to provide a fix or suggestion or say something? I
think the situation is quite series. Suggest to adjust the priority and have it
fixed quick. Thanks.
I've built current rawhide glibc for 8.0, in:
(particular changes are linuxthreads only (no NPTL), no INIT/FINI_ARRAY support,
no warnings about borken programs using errno, h_errno or _res symbols directly).
------- Additional Comment #11 From Richard Brown on 2003-01-26 04:42 -------
Also works on Redhat 7.0 running MySQL 3.23.54(rpm mysql-3.23.54a-3.70) using
glibc rpm version glibc-2.2.4-126.96.36.199
For the lazy you set a variable in my.cnf under [mysqld] section of file:
... and you don't even need to restart mysqld: ) Query once, then query again
and it takes effect.
Thank you Richard!!!! I would have to assume that the notes about mysql using
too large of a default thread stack for the latest glibc stack restrictions are
Probably not a redhat bug, more like a mysql/glibc configuration snafu. I would
like to add that recompiling mysql and/or glibc from source, alone, using
defaults, does not help one bit. I tried it on another box in my cube. This is
why I feel that it is not a redhat problem. However, redhat could help by
re-releasing the mysql rpm and adding this configuration parameter, provided
this fixes the issue on enough platforms/versions.
This certainly beats re-installing or upgrading the OS (unless your patches are
hopelessly out of date) which is how I fixed it last time this happened (after
an xinetd update on another RH7.1 box).
large bank (not allowed to say: )
An errata has been issued which should help the problem described in this bug report.
This report is therefore being closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files, please follow the link below. You may reopen
this bug report if the solution does not work for you.