[Xitami] Logfile issues

[Paul Reid]

 > ac%yy%mm%dd.log_000 Has anyone seen this before?

Yes, it means that a sub-process had to wait "too long" to get a 
write-lock on the log file (because another sub-process(es) was writing).

A faster drive or more buffering may help. However a random-request 
server has random spikes in traffic, and will always bind-up; you can 
only reduce the probability to very-small.

I have not dug in Xitami's guts in years. Can it pipe to a process? Can 
you write a process with far higher throughput than your O/S's 
writefile(append)? It is dubious because all modern O/Ses basically 
"write" to memory and buffer actual disk writes... they do what you want 
and have been highly optimized for someone's notion of typical 
transactional storage.

Multiple virtual hosts can write to multiple files, which may cut the bind.