Anybody using WD Blue drives for a storagenode?

but why… 1GB seems high for 1992, don’t tell me its a 5½"
lol thats awesome, i was just going to question why you would keep a drive like that… tho i think i do have a couple of old consumer grade drives around from that era… O.o
did speed testing on old drives recently… not been much development in hdd speeds since sata came into the picture… in fact some of the old 200gb drives actually beat the 1-2 tb drives… atleast in random io which was a bit surprising… or i think it was random… not really important lol at the time didn’t have an ssd to spare for the server. xD

who doesn’t like a drive with 700ms backlog… lol my drive being resilvered, the disk read speed floored at times, realized it was because my 4 drives was ofc ganging up on the 1 being resilvered, and with 48gb ram the “stall” actually lasts for a good while lol… not that it stalls… just the disk read floored…
the storage node seems totally fine with me processing around 200-300mb/s and wiriting 80-120mb/s
which is kinda odd because i can literally see when i turn on my windows server VM even tho it’s just idling doing basically nothing… might slightly affect ssd backlog tho…

Full Height too. I don’t use it anymore, but I recently checked it and there were no problems.

I doubt the few operations every 5 seconds or so would affect the performance that much. On the other hand, you could do something more complex and issue the dummy commands only if the drive load is below some opertaions/second value.

zfs resilver usually runs with very low priority and if there are other requests to the array the resilver/scrub process slows down to not impact the normal IO that much.

I was using touch in the crontab to prevent the hard disks from spinning down the platters, since they are connected to a Rock64 via USB.

Thanks to this thread I just found that one of the HDDs (WD Green) has a Load_Cycle_Count of 1547231, with 13239h of power on :slight_smile:

ioping seems to work fine for Blues and Greens, with the other node on a Red (24816h, 8763 load cycle) the crontab with touch seems sufficient.

1 Like