| Bug #84427 | mysqld_safe fails to start when using a fifo for log-error (regression) | ||
|---|---|---|---|
| Submitted: | 6 Jan 2017 2:38 | Modified: | 24 Feb 2017 1:48 |
| Reporter: | Travers Carter | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: Packaging | Severity: | S2 (Serious) |
| Version: | 5.6.35 | OS: | Linux (CentOS 7) |
| Assigned to: | Terje Røsten | CPU Architecture: | Any |
[6 Jan 2017 5:44]
Terje Røsten
Hi! Thanks for your report! A fix will be included in the next batch of releases.
[25 Jan 2017 21:11]
Jesse Duce
This happens on regular files as well when started with sudo. I normally run sudo service mysql start, which works fine on 5.6.34, but now, if the mysql_error.log doesn't exist beforehand, that command fails and now requires the service to be started as mysql. sudo -u mysql service mysql start. I will be skipping this patch in my test and production environments because of the potential for the service not to start on reboot.
[5 Feb 2017 3:17]
Phil Hildebrand
This also seems to happen on 5.5.54-38.6 on ubuntu. If the error.log does not exist, mysql will not start (same error). touching (creating one) the file fixes the issue
[24 Feb 2017 1:48]
Paul DuBois
Posted by developer: Noted in 5.5.55, 5.6.36, 5.7.18, 8.0.1 changelogs. mysqld_safe failed if the error log file named by the --log-error option was a FIFO.
Description: The changes made to mysqld_safe log file handling in MySQL 5.6.35 prevent mysqld from starting if the log-error points to a fifo rather than a regular file. mysqld safe complains: 170106 12:54:26 mysqld_safe error: log-error set to '/var/run/mysqld/log.fifo', however file don't exists. Create writable for user 'mysql'. However /var/run/mysqld/log.fifo does exist and is writeable by mysql, it is however a fifo and not a regular file. [root@dev conf.d]# ls -l /var/run/mysqld/log.fifo prw-r-----. 1 mysql mysql 0 Jan 6 13:05 /var/run/mysqld/log.fifo How to repeat: # Set up a fifo and listen on it mkfifo -m 0640 /var/run/mysqld/log.fifo chown mysql:mysql /var/run/mysqld/log.fifo cat /var/run/mysqld/log.fifo # Configure mysqld_safe to log to the fifo ----%<---- my.cnf ----%<---- [mysqld_safe] log-error = /var/run/mysqld/log.fifo ----%<--------%<--------%<---- # Start mysqld systemctl start mysqld.service Suggested fix: Extend the existence check in mysqld_safe to check for fifo (named pipe) files as well as regular files --- a/mysqld_safe 2017-01-06 13:34:21.237081735 +1100 +++ b/mysqld_safe 2017-01-06 13:32:41.941291759 +1100 @@ -633,7 +633,7 @@ fi fi - if [ -f "$err_log" ]; then # Log to err_log file + if [ -f "$err_log" -o -p "$err_log" ]; then # Log to err_log file log_notice "Logging to '$err_log'." elif [ "x$user" = "xroot" ]; then # running as root, mysqld can create log file; continue echo "Logging to '$err_log'." >&2