In the last installment in this series, I had made a 4x 8TB eSATA RAID5 set, and performance was slow, and it was unstable.
The disks went AWOL all the time with messages like
May 14 20:52:25 big-tex kernel: [ 3189.112594] ata5.03: hard resetting link ... May 14 20:53:32 big-tex kernel: [ 3255.728753] ata5.03: detaching (SCSI 4:3:0:0)
Things that could be causing this: 1) a bad disk enclosure. 2) a bad eSATA cable. 3) Bad port multiplier support in the chipset. 3) the disks. f) other.
So I switched out the easiest thing: I went from eSATA to USB3.
And it’s been rock stable since then. Dirt slow, but rock stable. Which is fitting.
How slow?
r b swpd free buff cache si so bi bo in cs us sy id wa 1 1 0 33320 1023516 2059076 0 0 0 42835 4476 7875 0 24 43 33 0 1 0 34432 1023596 2105640 0 0 0 40993 4341 7661 0 23 43 34 0 1 0 29632 1023684 2143344 0 0 1 39959 4268 7486 0 22 43 34
40MB/s write speed slow. That’s so slow that when I rip blu-ray disks, makemkv complains about the disk being too slow — it’s reading the disk faster than it can write to disk.
That’s depressing.
But Linux 4.1 was released today, and it’s supposed to have improved (i.e. faster) RAID5 support. I wasn’t holding my breath, but look:
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu---- r b swpd free buff cache si so bi bo in cs us sy id wa 2 1 0 30276 47320 3320332 0 0 123 47421 3952 6765 0 23 42 35 1 1 0 32604 47436 3300164 0 0 2 55326 4444 7456 0 28 39 33 0 1 0 29380 47564 3290044 0 0 2 53297 4302 7218 0 27 39 34 0 1 0 28660 47696 3276640 0 0 2 55374 4434 7387 0 28 39 34 0 1 0 28712 47840 3262992 0 0 8 56602 4460 7496 0 28 38 33 0 1 0 29992 47948 3252800 0 0 2 52579 4254 7102 0 26 39 35
It’s faster! It went from 40MB/s to 53MB/s. So it was somewhat CPU constrained? (It’s a dual core 1GHz AMD G-T40E processor…)
It’s still really slow, but perhaps that’s fast enough for makemkv…
What about with -o data=writeback?
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu---- r b swpd free buff cache si so bi bo in cs us sy id wa 0 2 0 31292 47952 3242020 0 0 2 58233 4377 7189 0 29 19 52 0 2 0 31144 48072 3231984 0 0 2 58205 4379 7118 0 29 18 52 0 2 0 28600 48184 3224956 0 0 2 57498 4282 6964 0 29 20 51 0 2 0 32928 48312 3210744 0 0 2 60983 4474 7302 0 32 19 50 1 1 0 30068 48460 3208108 0 0 15 62961 4593 7472 0 33 18 48
So about 58MB/s. Whee! The naughts called and wants their write speed back!
By the way, I connected my previous RAID box to the same machine with eSATA just to check whether the port multiplier stuff was unstable. And that’s been stable for weeks. So it is, as usual, certain combinations of certain gadgets via certain technology that’s the problem…