Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
G
gitlab-ce
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
0
Issues
0
List
Boards
Labels
Milestones
Merge Requests
1
Merge Requests
1
Analytics
Analytics
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Commits
Issue Boards
Open sidebar
nexedi
gitlab-ce
Commits
c5295f96
Commit
c5295f96
authored
Jun 26, 2016
by
Achilleas Pipinellis
Browse files
Options
Browse Files
Download
Plain Diff
Merge branch 'add-debugging-tips' into 'master'
Add debugging tips with gdb See merge request !4522
parents
6acdf1fc
67e83543
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
263 additions
and
0 deletions
+263
-0
doc/README.md
doc/README.md
+1
-0
doc/administration/troubleshooting/debug.md
doc/administration/troubleshooting/debug.md
+120
-0
doc/administration/troubleshooting/gdb-stuck-ruby.txt
doc/administration/troubleshooting/gdb-stuck-ruby.txt
+142
-0
No files found.
doc/README.md
View file @
c5295f96
...
...
@@ -44,6 +44,7 @@
-
[
Housekeeping
](
administration/housekeeping.md
)
Keep your Git repository tidy and fast.
-
[
GitLab Performance Monitoring
](
monitoring/performance/introduction.md
)
Configure GitLab and InfluxDB for measuring performance metrics.
-
[
Monitoring uptime
](
monitoring/health_check.md
)
Check the server status using the health check endpoint.
-
[
Debugging Tips
](
administration/troubleshooting/debug.md
)
Tips to debug problems when things go wrong
-
[
Sidekiq Troubleshooting
](
administration/troubleshooting/sidekiq.md
)
Debug when Sidekiq appears hung and is not processing jobs.
-
[
High Availability
](
administration/high_availability/README.md
)
Configure multiple servers for scaling or high availability.
-
[
Container Registry
](
administration/container_registry.md
)
Configure Docker Registry with GitLab.
...
...
doc/administration/troubleshooting/debug.md
0 → 100644
View file @
c5295f96
# Debugging Tips
Sometimes things don't work the way they should. Here are some tips on debugging issues out
in production.
## The GNU Project Debugger (gdb)
`gdb`
is a must-have tool for debugging issues. To install on Ubuntu/Debian:
```
sudo apt-get install gdb
```
On CentOS:
```
sudo yum install gdb
```
## Common Problems
Many of the tips to diagnose issues below apply to many different situations. We'll use one
concrete example to illustrate what you can do to learn what is going wrong.
### 502 Gateway Timeout after unicorn spins at 100% CPU
This error occurs when the Web server times out (default: 60 s) after not
hearing back from the unicorn worker. If the CPU spins to 100% while this in
progress, there may be something taking longer than it should.
To fix this issue, we first need to figure out what is happening. The
following tips are only recommended if you do NOT mind users being affected by
downtime. Otherwise skip to the next section.
1.
Load the problematic URL
1.
Run
`sudo gdb -p <PID>`
to attach to the unicorn process.
1.
In the gdb window, type:
```
call (void) rb_backtrace()
```
1.
This forces the process to generate a Ruby backtrace. Check
`/var/log/gitlab/unicorn/unicorn_stderr.log`
for the backtace. For example, you may see:
```ruby
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:33:in `block in start'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:33:in `loop'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:36:in `block (2 levels) in start'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:44:in `sample'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:68:in `sample_objects'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:68:in `each_with_object'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:68:in `each'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:69:in `block in sample_objects'
from /opt/gitlab/embedded/service/gitlab-rails/lib/gitlab/metrics/sampler.rb:69:in `name'
```
1.
To see the current threads, run:
```
apply all thread bt
```
1.
Once you're done debugging with
`gdb`
, be sure to detach from the process and exit:
```
detach
exit
```
Note that if the unicorn process terminates before you are able to run these
commands, gdb will report an error. To buy more time, you can always raise the
Unicorn timeout. For omnibus users, you can edit
`/etc/gitlab/gitlab.rb`
and
increase it from 60 seconds to 300:
```
ruby
unicorn
[
'worker_timeout'
]
=
300
```
For source installations, edit
`config/unicorn.rb`
.
[Reconfigure] GitLab for the changes to take effect.
[
Reconfigure
]:
../restart_gitlab.md#omnibus-gitlab-reconfigure
#### Troubleshooting without affecting other users
The previous section attached to a running unicorn process, and this may have
undesirable effects for users trying to access GitLab during this time. If you
are concerned about affecting others during a production system, you can run a
separate Rails process to debug the issue:
1.
Log in to your GitLab account.
1.
Copy the URL that is causing problems (e.g. https://gitlab.com/ABC).
1.
Obtain the private token for your user (Profile Settings -> Account).
1.
Bring up the GitLab Rails console. For omnibus users, run:
````
sudo gitlab-rails console
```
1.
At the Rails console, run:
```ruby
[1] pry(main)> app.get '<URL FROM STEP 1>/private_token?<TOKEN FROM STEP 2>'
```
For example:
```ruby
[1] pry(main)> app.get 'https://gitlab.com/gitlab-org/gitlab-ce/issues/1?private_token=123456'
```
1.
In a new window, run
`top`
. It should show this ruby process using 100% CPU. Write down the PID.
1.
Follow step 2 from the previous section on using gdb.
# More information
*
[
Debugging Stuck Ruby Processes
](
https://blog.newrelic.com/2013/04/29/debugging-stuck-ruby-processes-what-to-do-before-you-kill-9/
)
*
[
Cheatsheet of using gdb and ruby processes
](
gdb-stuck-ruby.txt
)
doc/administration/troubleshooting/gdb-stuck-ruby.txt
0 → 100644
View file @
c5295f96
# Here's the script I'll use to demonstrate - it just loops forever:
$ cat test.rb
#!/usr/bin/env ruby
loop do
sleep 1
end
# Now, I'll start the script in the background, and redirect stdout and stderr
# to /dev/null:
$ ruby ./test.rb >/dev/null 2>/dev/null &
[1] 1343
# Next, I'll grab the PID of the script (1343):
$ ps aux | grep test.rb
vagrant 1343 0.0 0.4 3884 1652 pts/0 S 14:42 0:00 ruby ./test.rb
vagrant 1345 0.0 0.2 4624 852 pts/0 S+ 14:42 0:00 grep --color=auto test.rb
# Now I start gdb. Note that I'm using sudo here. This may or may not be
# necessary in your setup. I'd try without sudo first, and fall back to adding
# it if the next step fails:
$ sudo gdb
GNU gdb (Ubuntu/Linaro 7.4-2012.04-0ubuntu2.1) 7.4-2012.04
Copyright (C) 2012 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. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-linux-gnu".
For bug reporting instructions, please see:
<http://bugs.launchpad.net/gdb-linaro/>.
# OK, now I'm in gdb, and I want to instruct it to attach to our Ruby process.
# I can do that using the 'attach' command, which takes a PID (the one we
# gathered above):
(gdb) attach 1343
Attaching to process 1343
Reading symbols from /opt/vagrant_ruby/bin/ruby...done.
Reading symbols from /lib/i386-linux-gnu/librt.so.1...(no debugging symbols found)...done.
Loaded symbols for /lib/i386-linux-gnu/librt.so.1
Reading symbols from /lib/i386-linux-gnu/libdl.so.2...(no debugging symbols found)...done.
Loaded symbols for /lib/i386-linux-gnu/libdl.so.2
Reading symbols from /lib/i386-linux-gnu/libcrypt.so.1...(no debugging symbols found)...done.
Loaded symbols for /lib/i386-linux-gnu/libcrypt.so.1
Reading symbols from /lib/i386-linux-gnu/libm.so.6...(no debugging symbols found)...done.
Loaded symbols for /lib/i386-linux-gnu/libm.so.6
Reading symbols from /lib/i386-linux-gnu/libc.so.6...(no debugging symbols found)...done.
Loaded symbols for /lib/i386-linux-gnu/libc.so.6
Reading symbols from /lib/i386-linux-gnu/libpthread.so.0...(no debugging symbols found)...done.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/i386-linux-gnu/libthread_db.so.1".
Loaded symbols for /lib/i386-linux-gnu/libpthread.so.0
Reading symbols from /lib/ld-linux.so.2...(no debugging symbols found)...done.
Loaded symbols for /lib/ld-linux.so.2
0xb770c424 in __kernel_vsyscall ()
# Great, now gdb is attached to the target process. If the step above fails, try
# going back and running gdb under sudo. The next thing I want to do is gather
# C-level backtraces from all threads in the process. The following command
# stands for 'thread apply all backtrace':
(gdb) t a a bt
Thread 1 (Thread 0xb74d76c0 (LWP 1343)):
#0 0xb770c424 in __kernel_vsyscall ()
#1 0xb75d7abd in select () from /lib/i386-linux-gnu/libc.so.6
#2 0x08069c56 in rb_thread_wait_for (time=...) at eval.c:11376
#3 0x080a20fd in rb_f_sleep (argc=1, argv=0xbf85f490) at process.c:1633
#4 0x0805e0e2 in call_cfunc (argv=0xbf85f490, argc=1, len=-1, recv=3075299660, func=0x80a20b0 <rb_f_sleep>)
at eval.c:5778
#5 rb_call0 (klass=3075304600, recv=3075299660, id=9393, oid=9393, argc=1, argv=0xbf85f490, body=0xb74c85a8, flags=2)
at eval.c:5928
#6 0x0805e35d in rb_call (klass=3075304600, recv=3075299660, mid=9393, argc=1, argv=0xbf85f490, scope=1,
self=<optimized out>) at eval.c:6176
#7 0x080651ec in rb_eval (self=3075299660, n=0xb74c4e1c) at eval.c:3521
#8 0x0805c31c in rb_yield_0 (val=6, self=3075299660, klass=<optimized out>, flags=0, avalue=0) at eval.c:5095
#9 0x0806a1e5 in loop_i () at eval.c:5227
#10 0x08058dbd in rb_rescue2 (b_proc=0x806a1c0 <loop_i>, data1=0, r_proc=0, data2=0) at eval.c:5491
#11 0x08058f28 in rb_f_loop () at eval.c:5252
#12 0x0805e0c1 in call_cfunc (argv=0x0, argc=0, len=0, recv=3075299660, func=0x8058ef0 <rb_f_loop>) at eval.c:5781
#13 rb_call0 (klass=3075304600, recv=3075299660, id=4121, oid=4121, argc=0, argv=0x0, body=0xb74d4dbc, flags=2)
at eval.c:5928
#14 0x0805e35d in rb_call (klass=3075304600, recv=3075299660, mid=4121, argc=0, argv=0x0, scope=1, self=<optimized out>)
at eval.c:6176
#15 0x080651ec in rb_eval (self=3075299660, n=0xb74c4dcc) at eval.c:3521
#16 0x080662c6 in rb_eval (self=3075299660, n=0xb74c4de0) at eval.c:3236
#17 0x08068ee4 in ruby_exec_internal () at eval.c:1654
#18 0x08068f24 in ruby_exec () at eval.c:1674
#19 0x0806b2cd in ruby_run () at eval.c:1684
#20 0x08053771 in main (argc=2, argv=0xbf860204, envp=0xbf860210) at main.c:48
# C backtraces are sometimes sufficient, but often Ruby backtraces are necessary
# for debugging as well. Ruby has a built-in function called rb_backtrace() that
# we can use to dump out a Ruby backtrace, but it prints to stdout or stderr
# (depending on your Ruby version), which might have been redirected to a file
# or to /dev/null (as in our example) when the process started up.
#
# To get aroundt this, we'll do a little trick and redirect the target process's
# stdout and stderr to the current TTY, so that any output from the process
# will appear directly on our screen.
# First, let's close the existing file descriptors for stdout and stderr
# (FD 1 and 2, respectively):
(gdb) call (void) close(1)
(gdb) call (void) close(2)
# Next, we need to figure out the device name for the current TTY:
(gdb) shell tty
/dev/pts/0
# OK, now we can pass the device name obtained above to open() and attach
# file descriptors 1 and 2 back to the current TTY with these calls:
(gdb) call (int) open("/dev/pts/0", 2, 0)
$1 = 1
(gdb) call (int) open("/dev/pts/0", 2, 0)
$2 = 2
# Finally, we call rb_backtrace() in order to dump the Ruby backtrace:
(gdb) call (void) rb_backtrace()
from ./test.rb:4:in `sleep'
from ./test.rb:4
from ./test.rb:3:in `loop'
from ./test.rb:3
# And here's how we get out of gdb. Once you've quit, you'll probably want to
# clean up the stuck process by killing it.
(gdb) quit
A debugging session is active.
Inferior 1 [process 1343] will be detached.
Quit anyway? (y or n) y
Detaching from program: /opt/vagrant_ruby/bin/ruby, process 1343
$
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment