Watchdog is down once every hour

[kentarosasaki]

I found out following error log in 1.2.5. This error log was appeared once every hour. In this gateway node, I don't use watchdog function. It means that I don't modify watchdog part in leo_gateway.conf.

[E]     [email protected]      2015-02-24 13:28:12.964202 +0900        1424752092      leo_watchdog:handle_info/2      119     {leo_watchdog_disk,{function_clause,[{leo_watchdog_disk,'-get_disk_data/2-fun-1-',[["/dev/mapper/VolGroup00-LV_root"]],[{file,"src/leo_watchdog_disk.erl"},{line,170}]},{leo_watchdog_disk,'-get_disk_data/2-lc$^1/1-0-',2,[{file,"src/leo_watchdog_disk.erl"},{line,178}]},{leo_watchdog_disk,check,4,[{file,"src/leo_watchdog_disk.erl"},{line,295}]},{leo_watchdog_disk,handle_call,2,[{file,"src/leo_watchdog_disk.erl"},{line,238}]},{leo_watchdog,handle_info,2,[{file,"src/leo_watchdog.erl"},{line,112}]},{gen_server,handle_msg,5,[{file,"gen_server.erl"},{line,604}]},{proc_lib,init_p_do_apply,3,[{file,"proc_lib.erl"},{line,239}]}]}}
[E]     [email protected]      2015-02-24 13:28:12.965045 +0900        1424752092      null:null       0       gen_server leo_watchdog_disk terminated with reason: no case clause matching ok in leo_watchdog:handle_info/2 line 120
[E]     [email protected]      2015-02-24 13:28:12.965708 +0900        1424752092      null:null       0       ["CRASH REPORT ",[80,114,111,99,101,115,115,32,"leo_watchdog_disk",32,119,105,116,104,32,"0",32,110,101,105,103,104,98,111,117,114,115,32,"exited",32,119,105,116,104,32,114,101,97,115,111,110,58,32,[["no case clause matching ","ok"," in ",[["leo_watchdog",58,"handle_info",47,"2"],[32,108,105,110,101,32,"120"]]]," in ",[["gen_server",58,"terminate",47,"6"],[32,108,105,110,101,32,"744"]]]]]
[E]     [email protected]      2015-02-24 13:28:12.966528 +0900        1424752092      null:null       0       Supervisor leo_watchdog_sup had child leo_watchdog_disk started with leo_watchdog_disk:start_link(["/"], [], 80, 100, 262144, 262144, 5, 5000000) at <0.23974.374> exit with reason no case clause matching ok in leo_watchdog:handle_info/2 line 120 in context child_terminated
[E]     [email protected]      2015-02-24 13:28:13.161113 +0900        1424752093      null:null       0       Error in process <0.28582.375> on node '[email protected]' with exit value: {{badmatch,{error,{error,{already_started,<0.1056.0>}}}},[{leo_watchdog_disk,'-init/1-fun-0-',1,[{file,"src/leo_watchdog_disk.erl"},{line,225}]}]}

[mocchira]

@kentarosasaki
Thank you for reporting this issue.
There are two issues.

• Get watchdog enabled on leo_gateway even if you disabled it in leo_gateway.conf
• Failed to parse df -lk command result

Since the first one is a bug, we will fix it.

Regarding the second one,
Since df -lk output on your env is not expected for us, please let us know df -lk output on your leo_gateway.
We will fix the parse logic for df -lk based on your result.

Our expected result on xnix environments.

ubuntu@ip-10-126-25-246:~/dev/leofs_client_tests/aws-sdk-ruby$ df -lk
Filesystem     1K-blocks    Used Available Use% Mounted on
/dev/xvda1      20496628 7637664  11920972  40% /
none                   4       0         4   0% /sys/fs/cgroup
udev             3824544      12   3824532   1% /dev
tmpfs             765948     356    765592   1% /run
none                5120       0      5120   0% /run/lock
none             3829736       0   3829736   0% /run/shm
none              102400       0    102400   0% /run/user
/dev/xvdb       30824956   45124  29207352   1% /mnt

[shuichiro-makigaki]

$ df -lk
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/mapper/VolGroup00-LV_root
                      26730364   3776520  21596020  15% /
tmpfs                  8167252         0   8167252   0% /dev/shm
/dev/sda1                99150     32442     61588  35% /boot
none                   8167252        12   8167240   1% /tmp
/dev/mapper/VolGroup01-LV_leofs
                     262098944 134873592 127225352  52% /leofs

Oh, Filesystem name is too long.

However, LVM makes these long filesystem name usually. (e.g. OpenStack Cinder uses 32 chars uuid, which is longer than this case.)

[kentarosasaki]

Additionally, In case clause which is in leo_watchdog.erl 120 step, handle_fail return just OK, but there are no processing after that.
https://github.com/leo-project/leo_watchdog/blob/develop/src/leo_watchdog.erl#L120
https://github.com/leo-project/leo_watchdog/blob/develop/src/leo_watchdog_disk.erl#L247

Of course, I'm not sure the detail around watchdog, but I got worried about the behaviour in case clause.

[mocchira]

@kentarosasaki no problem.
handle_fail ONLY can be used when there are something to be rollbacked.
if you write some code which have something side effects in handle_call,
you can use handle_fail to rollback this side effects.

[mocchira]

@shuichiro-makigaki Thanks.
I found out how to avoid this strange behaviour(when a filesystem name is too long).

http://git.savannah.gnu.org/gitweb/?p=coreutils.git;a=blob;f=src/df.c;h=4523c440f5f14f4933545841d54c68b525e97897;hb=ca637bff0e29732aad99f4c08afbf61c44ed94f0#l390

This behaviour has been fixed with gnu df version >= v8.10-40-g99679ff.

In case of lower versions,
Adding the posix_format option should fix this unexpected starting new line behaviour.
Please try df -lkP out on leo_gateway and please let us the result.
If this works for you, we will fix this issue by replacing df -lk with df -lkP.

[shuichiro-makigaki]

$ df --version
df (GNU coreutils) 8.4
Copyright (C) 2010 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Torbjorn Granlund, David MacKenzie, and Paul Eggert.
$ df -lkP
Filesystem         1024-blocks      Used Available Capacity Mounted on
/dev/mapper/VolGroup00-LV_root  26730364   7594808  17777732      30% /
tmpfs                  8167252         0   8167252       0% /dev/shm
/dev/sda1                99150     32442     61588      35% /boot
none                   8167252        20   8167232       1% /tmp
/dev/mapper/VolGroup01-LV_leofs 104816640     33000 104783640       1% /leofs

Looks good.

[mocchira]

Fixed with this commit.
leo-project/leo_watchdog@767897a

Confirmed CentOS 6.5/UbuntuLTS.
Now test is ongoing on other xunix platforms.

[mocchira]

Confirmed the latest FreeBSD/SmartOS.

@kentarosasaki @shuichiro-makigaki
This fix will be included in the next release( probably 1.2.7 ).

Thanks for you all contributions.