Buffalo Forums

Subject:  [Terastation Live] to [Linkstation Live] Backup is unreliable (HS-DHTGL/R5 2.14 to LS-CHL-V2B2F)

I have both devices on an ethernet LAN (same subnet).  Both with latest Firmware.

I have two backup jobs on the TS - one set to run in the morning and the other set to in the afternoon (otherwise jobs are the same).

When I inspect the logs, sometimes the backup jobs runs successfully, and sometimes fails to start (RSYNC failures).  For example, today the morning job failed and the afternoon one succeded.  Sometimes both succeed, sometimes both fail.  When it runs, it runs fine.

Both are on DHCP, but the IP addresses ahve not changed during this period.  I can always access admin pages of both devices. Sleep timers are disabled on the LS.

Any suggestions?  

[ Please note that I am getting "Connection refused" error sometimes and "Connection timeout" errors at other times - see below.]

OK, after trying different things, the problem appears solved using the following workaround:

                Set the 'Ethernet Frame Size' to 4102 (Jumbo) on both units.

- Setting it to the default 1518 seems to give the first error (Connection refused)

    -------- BEGIN BACKUP 2010/06/26 03:00:04
    rsync: failed to connect to 192.168.1.205: Connection refused (111)
    rsync error: error in socket IO (code 10) at clientserver.c(107) [sender=2.6.8]
    -------- END BACKUP 2010/06/26 03:00:04

- Setting it to the 9694 (Jumbo) seems to give the second error (Connection timeout)

-------- BEGIN BACKUP 2010/06/26 12:48:55
building file list ...
rsync: read error: Connection timed out (110)
rsync error: error in rsync protocol data stream (code 12) at io.c(614) [sender=2.6.8]
-------- END BACKUP 2010/06/26 13:04:59

I will watch and see in the coming days if this is a reliable workaround (I would not consider it a solution, just a workaround).

Thanks to references I found on the internet regarding changing MTUs.  

Hopefully someone from Buffalo can look into this issue further (TCP timers-related??)

Things are working reliably now, but I am not sure what fixed the problem - 1) the ethernet frame size change I noted above, or 2) moving the switch at the back to of the unit to the "fully on" position.

Well, with the LinkStation being on AUTO if you don't have NAS Navi installed on a PC on the network (XP) then the "Keep Alive" service won't be installed along with it.  After a certain amount of time if the LS doesn't receive packets from this service to "stay awake" it will hibernate until it receives that data.