Jordan Delong
363d1bb20f
This change is mostly for FB internal organizational reasons. Building is not effected beyond the fact that the target now lands in hphp/hhvm/hhvm rather than src/hhvm/hhvm.
28 linhas
1.0 KiB
Plaintext
28 linhas
1.0 KiB
Plaintext
|
|
<h2>How to Debug Server Problems</h2>
|
|
|
|
1. Crashing and other memory problems
|
|
|
|
Normally a crash generates /tmp/stacktrace.<pid>.log file that has stacktrace
|
|
where crash happens. Sometimes, a crash can happen badly without giving crash
|
|
log a chance to generate a readable stack. Then use ResourceLimit.CoreFileSize
|
|
to turn on coredumps to capture a crash.
|
|
|
|
Another way is to use Debug.RecordInput option to capture bad HTTP requests.
|
|
|
|
Debug.RecordInput = true
|
|
Debug.ClearInputOnSuccess = false
|
|
|
|
Then replay with "-m replay" option from the compiled program at command line.
|
|
This way, one can run it under valgrind to detect memory problems.
|
|
|
|
./program -m replay -c config.hdf --count=3 /tmp/hphp_request_captured
|
|
./program -m replay -c config.hdf captured_request1 captured_request2
|
|
./program -m replay -c config.hdf --count=2 req1 req2
|
|
|
|
2. Server hanging and other status problems
|
|
|
|
Admin server commands provide status information that may be useful for
|
|
debugging, esp. /status.json command that will tell network I/O status of each
|
|
thread.
|