On Sat, 2016-05-14 at 01:52 +1200, Amos Jeffries wrote: > The default action should be to fetch each range request separately > and > in parallel. Not caching the results. > > When admin has set only the range offset & quick-abort to force full > object retrieval the behaviour Heiler mentions happens - lots of > upstream bandwidth used for N copies. > The first one to complete starts to be used as a HIT for future > reuqests. But as each of the initial transfers completes it replaces > the > previously cached object as the one being hit on. > > So timing is critical, if Squid happens to delay any of the parallel > requests just long enough in its TCP accept() queue, auth or ACL > processing they could become HITs on an already finished object. > > Amos Yes, you was right! Timing is very critical. I've simulated two concurrent transfers (without 'collapsed_forwarding on' and with 'range_offset_limit none') using this code: --- #!/bin/bash export http_proxy="127.0.0.1:3128" curl --range $((1024 * 1024 * 1))-$((1024 * 1024 * 2)) http://mirror.co mnet.uz/centos/7/os/x86_64/images/efiboot.img > /dev/null & curl --range $((1024 * 1024 * 3))-$((1024 * 1024 * 4)) http://mirror.co mnet.uz/centos/7/os/x86_64/images/efiboot.img > /dev/null --- And got two MISS': 1463150025.340 987 127.0.0.1 TCP_MISS/206 1048943 GET http://mirror. comnet.uz/centos/7/os/x86_64/images/efiboot.img - HIER_DIRECT/91.196.76.102 application/octet-stream 1463150026.315 1963 127.0.0.1 TCP_MISS/206 1048943 GET http://mirror. comnet.uz/centos/7/os/x86_64/images/efiboot.img - HIER_DIRECT/91.196.76.102 application/octet-stream Then, I purged the object repeated with 'collapsed_forwarding on' and got MISS and HIT: 1463150169.010 370 127.0.0.1 TCP_MISS/206 1048943 GET http://mirror. comnet.uz/centos/7/os/x86_64/images/efiboot.img - HIER_DIRECT/91.196.76.102 application/octet-stream 1463150169.476 836 127.0.0.1 TCP_HIT/206 1048950 GET http://mirror.c omnet.uz/centos/7/os/x86_64/images/efiboot.img - HIER_NONE/- application/octet-stream Amos, thank you very much for the detailed explanation. collapsed_forwarding is really useful option for this situation. _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
