pt-stalk "--config" incorrect parsing

[Emanuelis]

# strace -e open pt-stalk --config=/tmp/test.conf 2>&1 | grep \\.conf
open("/etc/percona-toolkit/pt-stalk.conf", O_RDONLY) = 3

There are two bugs:
1. Documentation (http://www.percona.com/doc/percona-toolkit/2.2/pt-stalk.html#cmdoption-pt-stalk--config) doesn't say anything about default config files:
/etc/percona-toolkit/percona-toolkit.conf
/etc/percona-toolkit/$TOOL.conf
$HOME/.percona-toolkit.conf
$HOME/.$TOOL.conf

2. --config doesn't work with a "=" (see my example).

This unclarity/bug caused a crash of whole PXC cluster :(

pt-stalk 2.2.13

[Nilnandan Joshi (nilnandan-joshi)]

Hi,

For 1, there is already mentioned in documentation that where default config files can be. Check this
https://www.percona.com/doc/percona-toolkit/2.2/configuration_files.html#read-order

For 2, Unable to reproduce the error. Can you try to check with later version 2.2.14? Also is it possible for you to provide test case?

nilnandan@desktop:~$ sudo pt-stalk --no-stalk --iterations=2 --sleep=30 --config=/home/nilnandan/.my.cnf --dest=/home/nilnandan
2015_06_09_14_12_10 Starting /usr/bin/pt-stalk --function=status --variable=Threads_running --threshold=25 --match= --cycles=0 --interval=1 --iterations=2 --run-time=30 --sleep=30 --dest=/home/nilnandan --prefix= --notify-by-email= --log=/var/log/pt-stalk.log --pid=/var/run/pt-stalk.pid --plugin=
2015_06_09_14_12_10 Not stalking; collect triggered immediately
2015_06_09_14_12_10 Collect 1 triggered
2015_06_09_14_12_10 Collect 1 PID 24848
2015_06_09_14_14_01 Collect 1 done
2015_06_09_14_14_01 Sleeping 30 seconds after collect
2015_06_09_14_14_31 Not stalking; collect triggered immediately
2015_06_09_14_14_31 Collect 2 triggered
2015_06_09_14_14_31 Collect 2 PID 16965
2015_06_09_14_14_31 Collect 2 done
2015_06_09_14_14_31 Waiting up to 90 seconds for subprocesses to finish...
2015_06_09_14_15_02 Exiting because no more iterations
2015_06_09_14_15_02 /usr/bin/pt-stalk exit status 0
nilnandan@desktop:~$
nilnandan@desktop:~$ pt-stalk --version
pt-stalk 2.2.14
nilnandan@desktop:~$

[Emanuelis (emanuelis)]

For 1, I meant that there is no single word about configuration files read order in _pt-stalk_ documentation page. It would be valuable to add a link to "configuration files" page somewhere in https://www.percona.com/doc/percona-toolkit/2.2/pt-stalk.html#cmdoption-pt-stalk--config section.

For 2,
/usr/bin/pt-stalk contains lines:
 249 if [ $# -ge 2 ] && [ "$1" = "--config" ]; then
 250 shift # --config
 251 local user_config_files="$1"
 252 shift # that ^
 253 local IFS=","
 254 for user_config_file in $user_config_files; do
 255 _parse_config_files "$user_config_file"
 256 done
 257 else

So, it checks if the first argument is "--config":
[ "$1" = "--config" ]

When the first argument is "--config=file.conf" this doesn't work (so your example is not loading /home/nilnandan/.my.cnf file)

Now I can see in https://www.percona.com/doc/percona-toolkit/2.2/configuration_files.html#specifying that this behavior is documented globally for Percona Toolkit. But, please, add a link on every tool's documentation page near "--config".

[Nilnandan Joshi (nilnandan-joshi)]

Set to confirmed as a feature request. We should specify above details on respective pages.

[bar down (bardown)]

Are we sure this pt-stalk --config command actually parses correctly still?

Aside from the above mention of --config= not working, it seems things are still not parsed correctly (or honored) when used without the = sign.

For instance when using multiple configs, it seems that the LAST config is the only one honored:

pt-stalk --config ./deadlocks.conf,./longquery.conf,./highcnx.conf --dest=/var/lib/mysql/pt-stalk --log=/var/log/mysql/pt-stalk/pt-stalk.log --sleep=30 --defaults-file=/etc/percona-toolkit.conf

output:
2016_04_25_15_51_59 Starting /usr/bin/pt-stalk --function=status --variable=Threads_running --threshold=100 --match= --cycles=5 --interval=1 --iterations= --run-time=30 --sleep=30 --dest=/var/lib/mysql/pt-stalk --prefix= --notify-by-email= --log=/var/log/mysql/pt-stalk/pt-stalk.log --pid=/var/run/pt-stalk.pid --plugin=

In this example, my longquery.conf is not triggering, but if I put it as the last argument in the list of configs, it does:
[root@cer-lucastestl1 configs]# pt-stalk --config ./deadlocks.conf,./highcnx.conf,./longquery.conf --dest=/var/lib/mysql/pt-stalk --log=/var/log/mysql/pt-stalk/pt-stalk.log --sleep=30 --defaults-file=/etc/percona-toolkit.conf
2016_04_25_15_52_46 Starting /usr/bin/pt-stalk --function=/usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh --variable=Long_Query --threshold=0 --match= --cycles=5 --interval=1 --iterations= --run-time=30 --sleep=30 --dest=/var/lib/mysql/pt-stalk --prefix= --notify-by-email= --log=/var/log/mysql/pt-stalk/pt-stalk.log --pid=/var/run/pt-stalk.pid --plugin=
2016_04_25_15_53_52 Check results: /usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh(Long_Query)=1, matched=yes, cycles_true=1
2016_04_25_15_53_53 Check results: /usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh(Long_Query)=1, matched=yes, cycles_true=2
2016_04_25_15_53_54 Check results: /usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh(Long_Query)=1, matched=yes, cycles_true=3
2016_04_25_15_53_55 Check results: /usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh(Long_Query)=1, matched=yes, cycles_true=4
2016_04_25_15_53_56 Check results: /usr/local/jump/bin/mysql/pt-stalk/functions/longquery.sh(Long_Query)=1, matched=yes, cycles_true=5
2016_04_25_15_53_56 Collect 1 triggered
2016_04_25_15_53_56 Collect 1 PID 50385
2016_04_25_15_53_56 Collect 1 done
2016_04_25_15_53_56 Sleeping 30 seconds after collect

Version:
pt-stalk 2.2.7

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-1278