I am in process of using clamd as a service for scanning multiple files in parallel.
My workflow:
- ReadFile on client machine
- Stream data in blocks to clamav service
- Open clamd socket connection and write data on UnixSocket (clamd.ctl) (using clamd.ClamdUnixSocket())
This workflow works perfectly well when I have small files or just a few files being transferred over to my clamav service.
But as number of files in parallel increase, I am observing high write time on clam.ClamdUnixSocket.instream (internally, clamd.ctl unix socket).
Trying with 32-64 parallel files.
I also tried switching to network socket (TCP based) which also gives similar performance.
#req is HTTP request (python)
ipfile = req.stream
cus = clamd.ClamdUnixSocket()
scan_res = cus.instream(ipfile)
Questions:
- Is this a known issue ?
- Are there any configurations which can be tweaked for better performance.
ClamAV information:
clamav version: 0.103.8
clamd daemon is installed on linux server
Virus definitions are up to date.
