Just a short note here. Many Linux distributions by default uses the mq-deadline scheduler for HDDs. I’ve just found out that this scheduler only gained support for I/O classes in Linux 5.14 (August 2021), so this patch won’t improve anything immediately for users of e.g. some long-term supported distributions that use older kernels (like Debian Bullseye—thankfully, Bookworm is just around the corner!), or users of NASes that don’t get their kernels updated. It may be useful for node operators to explicitly check whether their I/O scheduler supports the Idle class, and maybe update the kernel, maybe switch to BFQ, which does support I/O classes.
My ReadyNAS is now discontinued and no further OS development is coming.
It is pretty closed down so no kernel upgrade will be possible.
It is based on Debian Jessie.
No, just the feature introduced in this commit will not be effective.
This is just an API call availability. Doesn’t mean it works in all possible circumstances. This is the commit that adds support to the mq-deadline scheduler.
Wait a minute!.. there are 2 Filewalkers?
What is one for and what is the other for?
Should we put both as “lazy”?
Man, my run command keeps getting bigger and bigger…
I’m on Synology too, but maybe it helps, maybe not, I just want to understand the options.
Already checked when you first posted. It mentioned something with deadline. Not at my computer ATM, but I can check the exact thing later if you want.
this version dont solve problem with
C:\Program Files\Storj1\Storage Node>storagenode.exe exit-status --identity-dir “C:\Identity1\storagenode” --log.output stderr --server.private-address 127.0.0.1:7779
2023-05-06T10:14:46.634+0300 INFO Anonymized tracing enabled {“Process”: “storagenode”}
2023-05-06T10:14:46.642+0300 FATAL Failed to load identity. {“Process”: “storagenode”, “error”: “file or directory not found: open \identity.cert: The system cannot find the file specified.”, “errorVerbose”: “file or directory not found: open \identity.cert: The system cannot find the file specified.\n\tstorj.io/common/identity.Config.Load:326\n\tmain.cmdGracefulExitStatus:186\n\tmain.newGracefulExitStatusCmd.func1:59\n\tstorj.io/private/process.cleanup.func1.4:399\n\tstorj.io/private/process.cleanup.func1:417\n\tgithub.com/spf13/cobra.(*Command).execute:852\n\tgithub.com/spf13/cobra.(*Command).ExecuteC:960\n\tgithub.com/spf13/cobra.(*Command).Execute:897\n\tstorj.io/private/process.ExecWithCustomOptions:113\n\tmain.main:29\n\truntime.main:250”}
Filewalker hell…
I hope all of them run as low prio. Because it sounds like none of them really needs to run with the same prio like customer up- and downloads.
This one would be great if it would go through the temp dir as well to get rid of leftover partial files that for some reason did not get deleted.
There’s too much trouble keeping FW on. I turned it off on all my nodes. I don’t see any differences in reported space. The DBs are keeping the scores flowlessly. If you let 10% free space as recomanded, you should not have any problems. FW hammers HDD, reducing it’s lifetime and loosing races, so you loose money from node going down, replacing HDD and lost races. I don’t have any good expectations for the lazy mode either; when ingress and egress peaks, you don’t have room for anything else taking HDD time, low or high priority. Why should you keep on an optional process that affects node’s performance and cost you money?
