High CPU usage, caching gem 0%, slow to LSP requests

[fishBone000]

Hello!
I'm trying to use solargraph on metasploitframework, but I noticed:

1. Indexing is fine. There's 2000 files to be indexed, the process took about 2 seconds.
2. But my editor keeps flashing "Caching gem", and the progress is always 0%. I have never seen any LSP status flashing, perhaps it's starting the caching process over and over again?
3. High CPU usage, around 70% (100% for one core, 200% for 2 cores)
4. Slow to LSP requests, e.g. it takes around 10 sec to respond "go to definition", takes whole damn while to get document symbols.

Environment:
OS: Arch Linux
solargraph: 0.55.1
editor: nvim
metasploit-framework: 6.4.68

Thx!

[fishBone000]

Problem resolved.
Reason is that I was using Ruby provided by system. Seems Solargraph is trying to do some caching inside system directory which requires root privilege.
I installed ruby in user folder (via rvm). Issues gone.

[apiology]

I've encountered something similar when my dev setup is goofy (e.g., having my bundler configuration suddenly be invalid while an LSP session is running. Might be worth doing an exponential backoff on error.

[fishBone000]

I'm actually not sure if it's the right solution.
Same issue appeared once after I started using rvm, but after I restart nvim things were fine again.

If someone encountered this issue with ruby installed in user dir, feel free to open an new issue. I might reopen it if I find it necessary.

[gongfarmer]

I am still experiencing this issue on solargraph 0.57.0, nvim 0.11.5.
It frequently gets into a loop where it is constantly flashing a Caching gem message.
This is on debian 13, ruby 3.3.8.

I have it currently in this state, here are some observations:

Neovim's :LspInfo command is reporting this on the buffer that is currently in this state:

vim.lsp: Active Clients ~
- solargraph (id: 1)
  - Version: ? (no serverInfo.version response)
  - Root directory: ~/git2/build
  - Command: { "solargraph", "stdio" }
  - Settings: {
      solargraph = {
        diagnostics = true
      }
    }
  - Attached buffers: 1

There is one core in usage at 100% and lots of solargraph processes are running according to top.

It appears that solargraph is constantly forking.
When I run ps aux | grep solargraph repeatedly, I notice that every run shows some different PIDs.

fhanson@fhanson-dev-2:~/git/sg-third-party (master)$ ps aux | grep [s]olargraph
fhanson     3893  0.9  0.7 1059876 275044 ?      Ssl  Dec18  16:03 /usr/bin/ruby3.3 /usr/local/bin/solargraph stdio
fhanson   364206  2.0  0.6 995252 258096 ?       Ssl  14:19   5:22 /usr/bin/ruby3.3 /usr/local/bin/solargraph stdio
fhanson   378285  0.1  0.5 964340 193652 ?       Ssl  17:29   0:05 /usr/bin/ruby3.3 /usr/local/bin/solargraph stdio
fhanson   378599  1.0  2.3 2062332 873344 ?      Ssl  17:31   0:48 /usr/bin/ruby3.3 /usr/local/bin/solargraph stdio
fhanson   392423  0.0  0.0  15316  9132 pts/9    T    18:36   0:00 /tmp/.mount_nvimFLGAFi/usr/bin/nvim /var/lib/gems/3.3.0/gems/solargraph-0.57.0/lib/solargraph/library.rb
fhanson   392429  0.0  0.0 1050004 16232 ?       Ssl  18:36   0:00 nvim /var/lib/gems/3.3.0/gems/solargraph-0.57.0/lib/solargraph/library.rb
fhanson   392434  0.9  0.1 160980 58256 ?        Ssl  18:36   0:05 /tmp/.mount_nvimFLGAFi/usr/bin/nvim --embed /var/lib/gems/3.3.0/gems/solargraph-0.57.0/lib/solargraph/library.rb
fhanson   392441  4.0  0.9 1682720 357932 ?      Ssl  18:36   0:20 /usr/bin/ruby3.3 /usr/local/bin/solargraph stdio
fhanson   397153  100  0.2 589804 99928 ?        Rl   18:45   0:01 /usr/bin/ruby3.3 /usr/local/bin/solargraph cache mail 2.8.1
fhanson   397178  100  0.0 470192 19948 ?        Rl   18:45   0:00 /usr/bin/ruby3.3 /usr/local/bin/solargraph cache rubocop 1.81.7

The number of PIDs is staying the same, but a few of them are always changing when I run this command once per second.