Replication between two servers incredibly slow

currently doing a (first) replication between two servers - both have 2 x 1Gb ethernet connections configured in bond ALB configuration attached to the same switch.

transferring a 2TB (used) share from one server to the other.

looking at the network performance it running at about (average) 12Mb/sec (Mbits - just claarify) and replication has been running for DAYS (3) not hours - this is crazy - i could have just copied them via an external ssd easier than using replication

any ideas as to why the replication is running so slowly?

i’ve tried to change the MTU to 9000 to enable jumbo frames between the two servers - it is not possible to do through the web interface

cwnd is set at 10 rttr is up around 10 - 18 - opensuse doesn’t give a lot of hints on optimising tcp

a 2TB transfer took longer than 4 days between two rockstor appliances on the same 1Gb switch in bonded ALB mode - somethings wrong with the network rockstor is setting up

Which version of Rockstor and Leap are you running on? And if you can check, which version of btrfs-progs?

I assume, if you’re doing a straight up file copy/move between both systems, the speed is much closer to saturating the network connection?

1 Like

A 2TB transfer using a single 1Gb connection should take about 9-11 hours.

Early on, I looked at a lot of aggregation techniques and discovered a myriad of glitches and problems with that kind of setup; particularly with budget components used in my home NAS/LAN setups. Ever lusting for speed, I ended up going with a 10G fiber setup that was relatively cheap and works flawlessly with Winderz and Linux.

If the ALB aggregation scheme doesn’t work, I’m sure it’s one of the tons of problems that a Goodgal search shows up. For now, I would skip the ALB thing and use a single 1G connection until you can fully detail the specs you need and/or wish to achieve. Going straight to a 2.5G or 10G setup may be the least problematic.

JMHO.

I wish you the very best of luck!!! And as always, have fun!

:sunglasses:

PS: A replication between my 2 NAS setups on 10G went about 295MBs… but I don’t use replication anymore.

4 Likes

I don’t know for sure, but I know if your switch doesn’t support link aggregation your bonded connection may not be working as anticipated and you might be dealing with packet collisions. Like @Tex1954 said your issue may be related to the bonding. Try using a single connection and see if things get better. I’m not sure what your hardware options are but you can get 2.5g USB adapters somewhat cheap and 2.5g routers aren’t too bad on the wallet. I want 10g but man things get expensive even with only 8 ports :frowning:

4 Likes

My setups consist of main NAS and a backup NAS and main Winderz workstation. I also have a testing NAS on 1G.

The whole setup is about $540 or so with the cables.

This is about as cheap as I could make it a few years ago… still working fine!

I needed only 3 10G ports, so the QNap switch was perfect.

:sunglasses:

2 Likes