Hello,
If a large amount of data is POST'ed to a process running mod_fcgid, the
Apache child uses an excessive amount of memory when processing it.
The client request is properly received and the following statement from
the documentation is true: "Once the amount of request body read from
the client exceeds FcgidMaxRequestInMem bytes, the remainder of the
request body will be stored in a temporary file."
The problem occurs when the temporary file is being sent to the FastCGI
handling process via its IPC socket. Here is the corresponding function
which sends the prepared "output_brigade" to the IPC socket:
https://gist.github.com/famzah/125a91971fba3450dd4926fe13e0ede6
The documentation of apr_bucket_read() clearly states that "if buckets
are read in a loop, and aren't deleted after being processed, the
potentially large bucket will slowly be converted into RAM resident heap
buckets. If the file is larger than available RAM, an out of memory
condition could be caused."
I need your guidance, in order to fix this properly. I've researched a
bit and see the following possible options to fix this:
1. Delete each bucket after sending it to the "ipc_handle". I've looked
through the call tree and the *output_brigade is last used by
proc_write_ipc(). Therefore, it should be safe to empty it while
being processed there.
2. Take the same approach as mod_http2, which handles FILE buckets in a
different way. Instead of using apr_bucket_read(), they process FILE
buckets by apr_file_read() and manage the data buffer manually. This
way the original *output_brigade won't be modified and automatically
split by apr_bucket_read(). This requires more coding work.
Best regards.
--Ivan
