Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Tilo Kaltenecker wrote:
>Hi,
>
>I have two 2.6.11 systems with LSI Logic SATA Raid Controller and drbd 0.7.11.
>When setup the first node as the primary node and sync it to the second node, it
>runs 40-50 MBit/sec over the gigabit-connection (crossover).
>I have tested the network and disk bandwidth by doing nfsmount and copy a big file
>and I get 600 MBit/sec.
>
>
Hi Tilo,
I might want to do two things:
1. Make sure you don't have the tcp bic congestion bug
2. Increase the MTU-size to 9000 on your replication nic's
The 1st point really had bitten me for 5 days where large transfers
where suddenly very slow.. while iperf(8) could obtain 900Mbps on the
eth1 (e1000) with 9000MTU.
Here's a bit of the kernel Changelog:
><davem at davemloft.net>
> [PATCH] Fix BIC congestion avoidance algorithm error
>
> Since BIC is the default congestion control algorithm
> enabled in every 2.6.x kernel out there, fixing errors
> in it becomes quite critical.
>
> A flaw in the loss handling caused it to not perform
> the binary search regimen of the BIC algorithm
> properly.
>
> The fix below from Stephen Hemminger has been heavily
> verified.
>
> [TCP]: BIC not binary searching correctly
>
> While redoing BIC for the split up version, I discovered that the existing
> 2.6.11 code doesn't really do binary search. It ends up being just a slightly
> modified version of Reno. See attached graphs to see the effect over simulated
> 1mbit environment.
>
> The problem is that BIC is supposed to reset the cwnd to the last loss value
> rather than ssthresh when loss is detected. The correct code (from the BIC
> TCP code for Web100) is in this patch.
>
> Signed-off-by: Stephen Hemminger <shemminger at osdl.org>
> Signed-off-by: David S. Miller <davem at davemloft.net>
> Signed-off-by: Chris Wright <chrisw at osdl.org>
> Signed-off-by: Greg Kroah-Hartman <gregkh at suse.de>
>
>
>
Off-topic: I just installed my 2nd High Available drbd cluster - it's
running very well using an Areca:
ARECA RAID: 64BITS PCI BUS DMA ADDRESSING SUPPORTED
scsi0 : ARECA ARC1220 PCI-EXPRESS 8 PORTS SATA RAID CONTROLLER (RAID6-ENGINE Inside)
Driver Version 1.20.00.07
Vendor: Areca Model: ARC-1220-VOL#00 Rev: R001
Type: Direct-Access ANSI SCSI revision: 03
SCSI device sda: 2929686528 512-byte hdwr sectors (1500000 MB)
on RedHat Enteprise 4 with Update 1 - kernel 2.6.9-11smp which gives
around 70MB/sec sustained writes over drbd0.
(iozone -s 512m -r 64k -t 8 -i 0 -i 1 has finished in 1min 25sec.)
Goodluck,
Leroy