No, we are only talking about 200W of extra power. The drives would be powered anyway through a JBOD.
200 more watt every 24 bays with your future setup. scale up!
we can split in another thread. âWhat should I do to manage 1000 nodes when BIG customers comes?â
I want to be positive. I want to think that one day storj need a lot of more stored data and everyone have to think how to scale up.
2 Likes
So⌠is the testing over? Paused?
Weâve heard theyâve hit their performance targets: so maybe now theyâre configuring the capacity-reservation stuff that will only run fast for a couple hours each day?
Well, that post you linked suggested the high throughput would continue, but easing off a bit throughout the day.
It seems to have stopped completely, thoughâŚ
(Gives one of my âlettuce nodesâ some time to complete its filewalker, mind you!) 
3 Likes
Did someone at Storj read your post and just turn things back on?!?! 
1 Like
Yesterday we saw a huge drop in healthy pieces from production sats. This could be the reason for the test stop⌠
Or its just the weekend and more tests are set for Monday.
Does it take a lot of time for the repair work to bring numbers back to normal?
Seems like it was just a hiccup
There is some weirdness with repairs. I have two nodes that failed 1-2 repairs each with âfile not foundâ with associated audit scores drop 0.03%.
Both nodes are on zfs volumes, scrubbed monthly, both systems are on a UPSes, shut down gracefully on power outages (node is sent SIGTERM).
Iâve tried searching logs for the piece that wasnât found â and did not find any evidence of it it being ever uploaded, but since I donât store logs forever that could have been some very old piece.
I donât know what to think of that.
@littleskunk do repair workers also work on a snapshot and is it possible the 1 hour wait for expiration deletion is too short for the repair workers to be up to date on expired pieces?
Tests are still ongoing for the weekend.
6 Likes
I deal with this in my day job and DESPISE it. I visit this forum because itâs a refreshing change from the daily grind. While sometimes chaotic, I greatly appreciate the candid and open responses.
4 Likes
Love this behavior. I believe at one point it was mentioned that node performance metric would be added to the dashboard. Just a thought from this comment that a hit rate of âdial downâ might add to that insight on the dashboard. Is something like this in the works? I would love something that simple as a SNO to compare when optimizing things.
4 Likes
This problem also hits those with 10 USD/month: as long as their nodes are vetted, theyâre hit with regular traffic, not a substitute.
You invested. Maybe not for Storj, but you did spend extra money on top of your ISPâs router.
Same here. Iâve started hitting bottlenecks that I canât fix easily anymore, first with the ISP router, then, a bit by surprise, by my HBA card. Though, I was planning to get a new desktop anyway, so at some point Iâll probably make my old one into a quite a bit better NAS than my N36L; had the foresight to get a pretty flexible mobo and case for my current desktop. I basically just need to buy a single component for maybe 30 USD. We will see, because a lot depends on how Storj rules/traffic/environment develop.
It is nice when they complete before another run starts.
My ZyXel Viva router works perfectly serving 100Mbit up/down during the test. I do not consider it an enterprise grade.
strange. usually have many lanes at 6gb⌠it should not be a problem. how did you find out?
I suggest to set retain concurrency to 1
1 Like
One of the affected nodesâ audit score is back to 100%. keeping an eye on the other.
1 Like
You can always just ignore him.
Youâre not coming across very nice, either.
3 Likes