search

anatol / smart.go

Pure-Go library to access drive's S.M.A.R.T. information
MIT License
171 stars 10 forks source link

too many open files error when using smart.go in gotop client #6

Closed rare-magma closed 2 years ago

rare-magma commented 2 years ago

I'm integrating smart.go with gotop (cli activity monitor) to gather info about the disks temperature in a machine in this PR: https://github.com/xxxserxxx/gotop/pull/221

Unfortunately after a few minutes, either smart.Open() or sm.ReadSMARTData() or sm.ReadSMART() end up causing a "too many open files" error. Any idea on what might be the cause for this issue?

anatol commented 2 years ago

It might a file descriptor leak somewhere in the code.

Could you please:

  1. make sure you use the latest version of smart.go library. There was a file descriptor leak fixed not that far ago
  2. run your tool with strace -f $yourtool that reproduces the descriptor leak and post the last few hundred lines of the output?
  3. provide exact steps to reproduce the problem on my side
anatol commented 2 years ago

Also when the gotop starts reporting "too many files" please run sudo lsof -p $(pidof gotop) . It shows all files opened by the process. And then post the output here.

anatol commented 2 years ago

I pushed a possible fix to ataattr branch. Please try the branch and let me know if you see any improvement with the file descriptor leak.

rare-magma commented 2 years ago

I encountered this problem when I was using github.com/anatol/smart.go v0.0.0-20220404164659-a19872369963 as seen in this PR: https://github.com/xxxserxxx/gotop/pull/221

To reproduce this issue it's necessary to run gotop in a machine with at least a SATA disk. Instructions here (https://github.com/xxxserxxx/gotop)

My machine specs: Fedora Silverblue 36.20220613.0 Linux kernel v5.17.13-300.fc36.x86_64 go version go1.18.2 linux/amd64 2 SATA SSDs 2 HDDs

I've executed gotop with the following arguments: sudo gotop --layout disk --rate 10s --color nord --

here's the output of sudo strace -f gotop: strace.txt

output of sudo lsof -p $(pidof gotop): lsof.txt

I've tried to reproduce the issue with the atattr branch and haven't managed to get the error so far. I'll try again tomorrow.

anatol commented 2 years ago

I ran a simple test with code similar to what you have in gotop and I do not see any fd leaks. Try my latest changes and let me know if you have any problems.

anatol commented 2 years ago

@rare-magma how did the fix work? Do you still see the fd leak?

rare-magma commented 2 years ago

@anatol I tried to reproduce again with your latest changes and wasn't successful so I believe the issue is solved. Thanks!