I am a happy user of mod_scgi.

For the in-house web apps we develop, concurrent overlapping requests involving SCGI have been a non-issue sofar. This would change for sure as soon as the web apps are exposed to the internet.

But in our custom SCGI backend resp. the application logic written in Lua, almost no provision for locking shared resources such as files open for writing exists sofar.

Thus my question: is nginx serializing concurrent SCGI requests for me, buffering subsequent requests as long as a pending request is not finished by the SCGI server/backend? Or do I have to bite the bullet and implement locking etc. in the face of concurrent requests?

Thanks a lot in advance, nginz_zurich