On Sun, 3 Feb 2019, Phillip Moore wrote:
> On Sun, Feb 3, 2019 at 1:30 PM Julian Anastasov <ja@xxxxxx> wrote:
> > The templates are synced but when it happens depends on the sysctl
> > vars. What is your configuration?
> > grep . /proc/sys/net/ipv4/vs/sync_*
> We are pretty much running defaults with respect to sync I believe.
> > /proc/sys/net/ipv4/vs/sync_qlen_max:383646
> > /proc/sys/net/ipv4/vs/sync_refresh_period:0
> > /proc/sys/net/ipv4/vs/sync_retries:0
> > /proc/sys/net/ipv4/vs/sync_sock_size:0
> > /proc/sys/net/ipv4/vs/sync_threshold:3 50
Yep, look like defaults.
> We've had some unrelated issues on tcp session setup for non-ftp sessions
> if the network is in such flux that the SYN and SYN-ACK land on different
> hosts. This is rare and we try to avoid this happening but I think the 3
> above means 3rd packet, so it wouldn't have synced with the SYN-ACK comes.
> I appreciate your feedback with these settings with respect to the
> persistence templates.
Not sure how similar is your setup but I remember for recent
thread where SH and MH algorithms can be used:
In short, you can use SH scheduler (for FTP '--sched-flags sh-port'
is not suitable). The sync_persist_mode flag comes with 3.11 kernel, not
sure whether it is present in your kernel. But may be you do not need
persistence if it was used just for FTP session purposes, the SH/MH
schedulers take care to forward both FTP commands and data to same
backend server because by default they schedule only by source address,
the source port is ignored.
Julian Anastasov <ja@xxxxxx>
Please read the documentation before posting - it's available at:
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users