unofficial mirror of meta@public-inbox.org
 help / color / mirror / Atom feed
* Issues with `lei` as non-root
@ 2023-03-28  1:00 Louis DeLosSantos
  2023-03-28  1:32 ` Eric Wong
  0 siblings, 1 reply; 9+ messages in thread
From: Louis DeLosSantos @ 2023-03-28  1:00 UTC (permalink / raw)
  To: meta

Hello,

I'm experimenting with `lei` as a nice search tool for `lore.kernel.org`

Everything works fine with the caveat that it seems to break if I'm not root.

When using `lei` as non-root we get this error:

```
E: Linux::Inotify2->new: Too many open files at
/usr/share/perl5/vendor_perl/PublicInbox/DirIdle.pm line 40.
connect(/run/user/1000/lei/5.seq.sock): Connection refused (after
attempted daemon start)
```

Any ideas why this may occur? Is `lei` designed to only be ran as root
or is Fedora installing perl in an odd fashion which results in root
needing to be used?

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  1:00 Issues with `lei` as non-root Louis DeLosSantos
@ 2023-03-28  1:32 ` Eric Wong
  2023-03-28  1:36   ` public-inbox.org downtime expected in the next 6-8 hours Eric Wong
  2023-03-28  2:30   ` Issues with `lei` as non-root Louis DeLosSantos
  0 siblings, 2 replies; 9+ messages in thread
From: Eric Wong @ 2023-03-28  1:32 UTC (permalink / raw)
  To: Louis DeLosSantos; +Cc: meta

Louis DeLosSantos <louis.delos@gmail.com> wrote:
> Hello,
> 
> I'm experimenting with `lei` as a nice search tool for `lore.kernel.org`
> 
> Everything works fine with the caveat that it seems to break if I'm not root.
> 
> When using `lei` as non-root we get this error:

I've never used lei as root nor has any part of public-inbox
ever been intended to run as root.

> ```
> E: Linux::Inotify2->new: Too many open files at
> /usr/share/perl5/vendor_perl/PublicInbox/DirIdle.pm line 40.
> connect(/run/user/1000/lei/5.seq.sock): Connection refused (after
> attempted daemon start)
> ```
> 
> Any ideas why this may occur? Is `lei` designed to only be ran as root
> or is Fedora installing perl in an odd fashion which results in root
> needing to be used?

What's the output of `ulimit -n` and `lsof -p $(lei daemon-pid)`?

(you may need to use `ps -ef |grep lei-daemon` to get the PID
if lei is broken and using too many FDs, though)

`ulimit -n' is the open file limit, typically 1024 or higher.

If `lsof -p $PID` may reveal a bug in lei which leaves too many
files open.  lei (especially with inotify on Linux) should use
far less than 1024.

(FreeBSD may end up using far more open files, but that's a
different story)

^ permalink raw reply	[flat|nested] 9+ messages in thread

* public-inbox.org downtime expected in the next 6-8 hours
  2023-03-28  1:32 ` Eric Wong
@ 2023-03-28  1:36   ` Eric Wong
  2023-03-28  2:30   ` Issues with `lei` as non-root Louis DeLosSantos
  1 sibling, 0 replies; 9+ messages in thread
From: Eric Wong @ 2023-03-28  1:36 UTC (permalink / raw)
  To: meta; +Cc: Louis DeLosSantos

Btw, I need to finish off a HW downgrade due to increasing
hosting costs.  So retry sending any messages if you get bounces
or if your message doesn't show up on
https://public-inbox.org/meta/ after a few minutes.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  1:32 ` Eric Wong
  2023-03-28  1:36   ` public-inbox.org downtime expected in the next 6-8 hours Eric Wong
@ 2023-03-28  2:30   ` Louis DeLosSantos
  2023-03-28  2:52     ` Eric Wong
  1 sibling, 1 reply; 9+ messages in thread
From: Louis DeLosSantos @ 2023-03-28  2:30 UTC (permalink / raw)
  To: Eric Wong; +Cc: meta

> What's the output of `ulimit -n` and `lsof -p $(lei daemon-pid)`?
🖳  ulimit -n
1024

🖳  ps -ef |grep lei-daemon
root      861005       1  0 19:24 ?        00:00:00 lei-daemon
/tmp/lei-0/5.seq.sock
louis    1025477 1015489  0 22:22 pts/6    00:00:00 grep --color=auto lei-daemon

~
🖳 sudo  lsof -p 861005
COMMAND      PID USER   FD      TYPE             DEVICE SIZE/OFF    NODE NAME
lei-daemo 861005 root  cwd       DIR               0,34      246
2902791 /home/louis/Mail/linux-bpf
lei-daemo 861005 root  rtd       DIR               0,34      158     256 /
lei-daemo 861005 root  txt       REG               0,34    15984
746278 /usr/bin/perl
lei-daemo 861005 root  mem       REG               0,32
746278 /usr/bin/perl (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
669987 /usr/lib64/libstdc++.so.6.0.30 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
654072 /usr/lib64/libgcc_s-12-20221121.so.1 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
707780 /usr/lib64/libxapian.so.30.12.1 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
744976 /usr/lib64/perl5/vendor_perl/auto/Data/Dumper/Dumper.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2847838 /usr/lib64/perl5/vendor_perl/auto/Search/Xapian/Xapian.so
(path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
670001 /usr/lib64/libz.so.1.2.12 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
670059 /usr/lib64/libsqlite3.so.0.8.6 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
1142922 /usr/lib/locale/locale-archive (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
746285 /usr/lib64/perl5/auto/Sys/Hostname/Hostname.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
670032 /usr/lib64/libuuid.so.1.3.0 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2846948 /usr/lib64/perl5/vendor_perl/auto/Compress/Raw/Zlib/Zlib.so
(path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2848056 /usr/lib64/perl5/vendor_perl/auto/DBD/SQLite/SQLite.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
748851 /usr/lib64/perl5/vendor_perl/auto/DBI/DBI.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745669 /usr/lib64/perl5/auto/attributes/attributes.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2847712 /usr/lib64/perl5/vendor_perl/auto/Cpanel/JSON/XS/XS.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2848586 /usr/lib64/perl5/vendor_perl/auto/Linux/Inotify2/Inotify2.so
(path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2849070 /usr/lib64/perl5/vendor_perl/auto/Socket/MsgHdr/MsgHdr.so
(path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2847193 /usr/lib64/perl5/vendor_perl/auto/Sys/Syslog/Syslog.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745481 /usr/lib64/perl5/vendor_perl/auto/MIME/Base64/Base64.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745670 /usr/lib64/perl5/auto/re/re.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745506 /usr/lib64/perl5/vendor_perl/auto/Storable/Storable.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745611 /usr/lib64/perl5/vendor_perl/auto/Encode/Encode.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2849930 /usr/lib64/perl5/vendor_perl/auto/Email/Address/XS/XS.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
1143694 /usr/lib64/libm.so.6 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
1143691 /usr/lib64/libc.so.6 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745650 /usr/lib64/libperl.so.5.36.0 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2846914 /usr/lib64/perl5/vendor_perl/auto/Digest/SHA/SHA.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745663 /usr/lib64/perl5/auto/File/Glob/Glob.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745449 /usr/lib64/perl5/auto/IO/IO.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745355 /usr/lib64/perl5/vendor_perl/auto/Socket/Socket.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745492 /usr/lib64/perl5/vendor_perl/auto/List/Util/Util.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745259 /usr/lib64/perl5/auto/POSIX/POSIX.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
1019649 /usr/lib64/libcrypt.so.2.0.0 (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745556 /usr/lib64/perl5/vendor_perl/auto/Cwd/Cwd.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
2846928 /usr/lib64/perl5/vendor_perl/auto/Time/HiRes/HiRes.so (path
dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
745427 /usr/lib64/perl5/auto/Fcntl/Fcntl.so (path dev=0,34)
lei-daemo 861005 root  mem       REG               0,32
1143688 /usr/lib64/ld-linux-x86-64.so.2 (path dev=0,34)
lei-daemo 861005 root    0u      REG               0,38        0
3448 /tmp/lei-0/errors.log
lei-daemo 861005 root    1u      REG               0,38        0
3448 /tmp/lei-0/errors.log
lei-daemo 861005 root    2u      REG               0,38        0
3448 /tmp/lei-0/errors.log
lei-daemo 861005 root    3u  a_inode               0,14        0
15364 [eventpoll:4,5,7,8,14]
lei-daemo 861005 root    4u     unix 0x0000000033b859c6      0t0
4095008 /tmp/lei-0/5.seq.sock type=SEQPACKET (LISTEN)
lei-daemo 861005 root    5u     unix 0x000000007a56c532      0t0
4085549 type=SEQPACKET (CONNECTED)
lei-daemo 861005 root    6u     unix 0x00000000c6d0e27b      0t0
4085550 type=SEQPACKET (CONNECTED)
lei-daemo 861005 root    7r  a_inode               0,14        0   15364 inotify
lei-daemo 861005 root    8u  a_inode               0,14        0
15364 [signalfd]
lei-daemo 861005 root   14r     FIFO               0,13      0t0 4098533 pipe
lei-daemo 861005 root   16u     unix 0x00000000557447d0      0t0
4098534 type=SEQPACKET (CONNECTED)
lei-daemo 861005 root   17u     unix 0x00000000f60edd15      0t0
4098535 type=SEQPACKET (CONNECTED)

Above is 54 open sockets. Which seems fine.


Should daemon be running as root, if I intend to only use lei as user?

On Mon, Mar 27, 2023 at 9:32 PM Eric Wong <e@80x24.org> wrote:
>
> Louis DeLosSantos <louis.delos@gmail.com> wrote:
> > Hello,
> >
> > I'm experimenting with `lei` as a nice search tool for `lore.kernel.org`
> >
> > Everything works fine with the caveat that it seems to break if I'm not root.
> >
> > When using `lei` as non-root we get this error:
>
> I've never used lei as root nor has any part of public-inbox
> ever been intended to run as root.
>
> > ```
> > E: Linux::Inotify2->new: Too many open files at
> > /usr/share/perl5/vendor_perl/PublicInbox/DirIdle.pm line 40.
> > connect(/run/user/1000/lei/5.seq.sock): Connection refused (after
> > attempted daemon start)
> > ```
> >
> > Any ideas why this may occur? Is `lei` designed to only be ran as root
> > or is Fedora installing perl in an odd fashion which results in root
> > needing to be used?
>
> What's the output of `ulimit -n` and `lsof -p $(lei daemon-pid)`?
>
> (you may need to use `ps -ef |grep lei-daemon` to get the PID
> if lei is broken and using too many FDs, though)
>
> `ulimit -n' is the open file limit, typically 1024 or higher.
>
> If `lsof -p $PID` may reveal a bug in lei which leaves too many
> files open.  lei (especially with inotify on Linux) should use
> far less than 1024.
>
> (FreeBSD may end up using far more open files, but that's a
> different story)

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  2:30   ` Issues with `lei` as non-root Louis DeLosSantos
@ 2023-03-28  2:52     ` Eric Wong
  2023-03-28  3:05       ` Louis DeLosSantos
  0 siblings, 1 reply; 9+ messages in thread
From: Eric Wong @ 2023-03-28  2:52 UTC (permalink / raw)
  To: Louis DeLosSantos; +Cc: meta

Louis DeLosSantos <louis.delos@gmail.com> wrote:

<snip>

> Above is 54 open sockets. Which seems fine.

Actually, only 18 (0..17).  The `mem' stuff is mmap-ed and
doesn't count against `ulimit -n` (RLIMIT_NOFILE).

> Should daemon be running as root, if I intend to only use lei as user?

Definitely not; the lei-daemon is per-user.

I also forgot, inotify has its own per-user limits; perhaps
you're hitting those?

# show system-wide limits
$ head /proc/sys/fs/inotify/max_*

# show per-user inotify FDs (-nP speeds up lsof by avoiding lookups)
$ lsof -nP -u $USER |grep inotify

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  2:52     ` Eric Wong
@ 2023-03-28  3:05       ` Louis DeLosSantos
  2023-03-28  3:38         ` Eric Wong
  0 siblings, 1 reply; 9+ messages in thread
From: Louis DeLosSantos @ 2023-03-28  3:05 UTC (permalink / raw)
  To: Eric Wong; +Cc: meta

> Actually, only 18 (0..17).  The `mem' stuff is mmap-ed and
doesn't count against `ulimit -n` (RLIMIT_NOFILE).

Yup, you are right, did a quick "wc -l" without thinking about it.

> Definitely not; the lei-daemon is per-user.

Okay, maybe this is the issue to begin with? I installed lei from dnf.
I'm not sure what launches the daemon, is it launched on first run?

If that is the case, it was probably launched when I restored to `sudo
lei q ....` command.
But, if its running as systemd service, I could move it to user service.

# show system-wide limits
~
🖳  head /proc/sys/fs/inotify/max_*
==> /proc/sys/fs/inotify/max_queued_events <==
16384

==> /proc/sys/fs/inotify/max_user_instances <==
128

==> /proc/sys/fs/inotify/max_user_watches <==
524288

# show per-user inotify FDs (-nP speeds up lsof by avoiding lookups)
systemd      1686 louis    6r  a_inode               0,14         0
  15364 inotify
systemd      1686 louis   11r  a_inode               0,14         0
  15364 inotify
systemd      1686 louis   13r  a_inode               0,14         0
  15364 inotify
dbus-brok    1727 louis    7r  a_inode               0,14         0
  15364 inotify
swaync       1868 louis   15r  a_inode               0,14         0
  15364 inotify
dbus-brok    1898 louis    7r  a_inode               0,14         0
  15364 inotify
xdg-deskt    1914 louis   10r  a_inode               0,14         0
  15364 inotify
xdg-deskt    1938 louis   12r  a_inode               0,14         0
  15364 inotify
xdg-deskt    1968 louis   14r  a_inode               0,14         0
  15364 inotify
wireplumb    1979 louis   19r  a_inode               0,14         0
  15364 inotify
wireplumb    1979 louis   28r  a_inode               0,14         0
  15364 inotify
wireplumb    1979 louis   30r  a_inode               0,14         0
  15364 inotify
code         2362 louis   63r  a_inode               0,14         0
  15364 inotify
code         2362 louis   86r  a_inode               0,14         0
  15364 inotify
code         2362 louis  109r  a_inode               0,14         0
  15364 inotify
code         2456 louis   21r  a_inode               0,14         0
  15364 inotify
code         2493 louis   51r  a_inode               0,14         0
  15364 inotify
code         2547 louis   49r  a_inode               0,14         0
  15364 inotify
code         2547 louis   54r  a_inode               0,14         0
  15364 inotify
firefox      4168 louis   80r  a_inode               0,14         0
  15364 inotify
cgroupify    4347 louis    6r  a_inode               0,14         0
  15364 inotify
flatpak-s   53051 louis    7r  a_inode               0,14         0
  15364 inotify
obsidian    53073 louis   69r  a_inode               0,14         0
  15364 inotify
flatpak-p   53080 louis    7r  a_inode               0,14         0
  15364 inotify
obsidian    53135 louis   21r  a_inode               0,14         0
  15364 inotify
obsidian    53141 louis   56r  a_inode               0,14         0
  15364 inotify
tail       235359 louis    4r  a_inode               0,14         0
  15364 inotify
tail       251716 louis    4r  a_inode               0,14         0
  15364 inotify
tail       251821 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252069 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252266 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252399 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252434 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252749 louis    4r  a_inode               0,14         0
  15364 inotify
tail       252953 louis    4r  a_inode               0,14         0
  15364 inotify
tail       253611 louis    4r  a_inode               0,14         0
  15364 inotify
tail       253772 louis    4r  a_inode               0,14         0
  15364 inotify
tail       253858 louis    4r  a_inode               0,14         0
  15364 inotify
tail       253932 louis    4r  a_inode               0,14         0
  15364 inotify
tail       254159 louis    4r  a_inode               0,14         0
  15364 inotify
tail       254323 louis    4r  a_inode               0,14         0
  15364 inotify
tail       254425 louis    4r  a_inode               0,14         0
  15364 inotify
tail       255070 louis    4r  a_inode               0,14         0
  15364 inotify
tail       255592 louis    4r  a_inode               0,14         0
  15364 inotify
tail       256816 louis    4r  a_inode               0,14         0
  15364 inotify
tail       256939 louis    4r  a_inode               0,14         0
  15364 inotify
tail       257302 louis    4r  a_inode               0,14         0
  15364 inotify
tail       257435 louis    4r  a_inode               0,14         0
  15364 inotify
tail       257746 louis    4r  a_inode               0,14         0
  15364 inotify
tail       258071 louis    4r  a_inode               0,14         0
  15364 inotify
tail       258169 louis    4r  a_inode               0,14         0
  15364 inotify
tail       258279 louis    4r  a_inode               0,14         0
  15364 inotify
tail       258413 louis    4r  a_inode               0,14         0
  15364 inotify
tail       258715 louis    4r  a_inode               0,14         0
  15364 inotify
tail       259167 louis    4r  a_inode               0,14         0
  15364 inotify
tail       259289 louis    4r  a_inode               0,14         0
  15364 inotify
tail       259484 louis    4r  a_inode               0,14         0
  15364 inotify
tail       259623 louis    4r  a_inode               0,14         0
  15364 inotify
tail       259876 louis    4r  a_inode               0,14         0
  15364 inotify
tail       260275 louis    4r  a_inode               0,14         0
  15364 inotify
tail       260412 louis    4r  a_inode               0,14         0
  15364 inotify
tail       260518 louis    4r  a_inode               0,14         0
  15364 inotify
tail       260650 louis    4r  a_inode               0,14         0
  15364 inotify
tail       260818 louis    4r  a_inode               0,14         0
  15364 inotify
tail       261665 louis    4r  a_inode               0,14         0
  15364 inotify
tail       262636 louis    4r  a_inode               0,14         0
  15364 inotify
tail       262963 louis    4r  a_inode               0,14         0
  15364 inotify
tail       263781 louis    4r  a_inode               0,14         0
  15364 inotify
tail       264183 louis    4r  a_inode               0,14         0
  15364 inotify
tail       264242 louis    4r  a_inode               0,14         0
  15364 inotify
tail       264484 louis    4r  a_inode               0,14         0
  15364 inotify
tail       264769 louis    4r  a_inode               0,14         0
  15364 inotify
tail       264970 louis    4r  a_inode               0,14         0
  15364 inotify
tail       265488 louis    4r  a_inode               0,14         0
  15364 inotify
tail       265939 louis    4r  a_inode               0,14         0
  15364 inotify
tail       266145 louis    4r  a_inode               0,14         0
  15364 inotify
tail       267183 louis    4r  a_inode               0,14         0
  15364 inotify
tail       267352 louis    4r  a_inode               0,14         0
  15364 inotify
tail       267951 louis    4r  a_inode               0,14         0
  15364 inotify
tail       268135 louis    4r  a_inode               0,14         0
  15364 inotify
tail       268320 louis    4r  a_inode               0,14         0
  15364 inotify
tail       268892 louis    4r  a_inode               0,14         0
  15364 inotify
tail       269418 louis    4r  a_inode               0,14         0
  15364 inotify
tail       269631 louis    4r  a_inode               0,14         0
  15364 inotify
tail       339541 louis    4r  a_inode               0,14         0
  15364 inotify
tail       339854 louis    4r  a_inode               0,14         0
  15364 inotify
tail       340349 louis    4r  a_inode               0,14         0
  15364 inotify
tail       340566 louis    4r  a_inode               0,14         0
  15364 inotify
tail       340712 louis    4r  a_inode               0,14         0
  15364 inotify
tail       340858 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341122 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341356 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341499 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341643 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341858 louis    4r  a_inode               0,14         0
  15364 inotify
tail       341948 louis    4r  a_inode               0,14         0
  15364 inotify
tail       342271 louis    4r  a_inode               0,14         0
  15364 inotify
tail       342410 louis    4r  a_inode               0,14         0
  15364 inotify
tail       342557 louis    4r  a_inode               0,14         0
  15364 inotify
tail       342802 louis    4r  a_inode               0,14         0
  15364 inotify
tail       342944 louis    4r  a_inode               0,14         0
  15364 inotify
tail       343409 louis    4r  a_inode               0,14         0
  15364 inotify
tail       343690 louis    4r  a_inode               0,14         0
  15364 inotify
tail       343864 louis    4r  a_inode               0,14         0
  15364 inotify
tail       344049 louis    4r  a_inode               0,14         0
  15364 inotify
tail       344285 louis    4r  a_inode               0,14         0
  15364 inotify
tail       344429 louis    4r  a_inode               0,14         0
  15364 inotify
tail       345251 louis    4r  a_inode               0,14         0
  15364 inotify
tail       345541 louis    4r  a_inode               0,14         0
  15364 inotify
tail       345972 louis    4r  a_inode               0,14         0
  15364 inotify
tail       346163 louis    4r  a_inode               0,14         0
  15364 inotify
tail       346755 louis    4r  a_inode               0,14         0
  15364 inotify
tail       347063 louis    4r  a_inode               0,14         0
  15364 inotify
tail       350273 louis    4r  a_inode               0,14         0
  15364 inotify
tail       350316 louis    4r  a_inode               0,14         0
  15364 inotify
tail       350781 louis    4r  a_inode               0,14         0
  15364 inotify
tail       350942 louis    4r  a_inode               0,14         0
  15364 inotify
tail       351111 louis    4r  a_inode               0,14         0
  15364 inotify
tail       351746 louis    4r  a_inode               0,14         0
  15364 inotify
tail       353489 louis    4r  a_inode               0,14         0
  15364 inotify
tail       353801 louis    4r  a_inode               0,14         0
  15364 inotify
tail       354352 louis    4r  a_inode               0,14         0
  15364 inotify
tail       354638 louis    4r  a_inode               0,14         0
  15364 inotify
tail       354923 louis    4r  a_inode               0,14         0
  15364 inotify
tail       362029 louis    4r  a_inode               0,14         0
  15364 inotify
tail       367093 louis    4r  a_inode               0,14         0
  15364 inotify
waybar     811099 louis   37r  a_inode               0,14         0
  15364 inotify
waybar     811099 louis   38r  a_inode               0,14         0
  15364 inotify

On Mon, Mar 27, 2023 at 10:52 PM Eric Wong <e@80x24.org> wrote:
>
> Louis DeLosSantos <louis.delos@gmail.com> wrote:
>
> <snip>
>
> > Above is 54 open sockets. Which seems fine.
>
> Actually, only 18 (0..17).  The `mem' stuff is mmap-ed and
> doesn't count against `ulimit -n` (RLIMIT_NOFILE).
>
> > Should daemon be running as root, if I intend to only use lei as user?
>
> Definitely not; the lei-daemon is per-user.
>
> I also forgot, inotify has its own per-user limits; perhaps
> you're hitting those?
>
> # show system-wide limits
> $ head /proc/sys/fs/inotify/max_*
>
> # show per-user inotify FDs (-nP speeds up lsof by avoiding lookups)
> $ lsof -nP -u $USER |grep inotify

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  3:05       ` Louis DeLosSantos
@ 2023-03-28  3:38         ` Eric Wong
  2023-03-28  4:08           ` Louis DeLosSantos
  0 siblings, 1 reply; 9+ messages in thread
From: Eric Wong @ 2023-03-28  3:38 UTC (permalink / raw)
  To: Louis DeLosSantos; +Cc: meta

Louis DeLosSantos <louis.delos@gmail.com> wrote:
> > Definitely not; the lei-daemon is per-user.
> 
> Okay, maybe this is the issue to begin with? I installed lei from dnf.
> I'm not sure what launches the daemon, is it launched on first run?
> 
> If that is the case, it was probably launched when I restored to `sudo
> lei q ....` command.
> But, if its running as systemd service, I could move it to user service.

You shouldn't need to manage it as a service; it's auto-started
and killing it is harmless in most cases.  I'm considering it
have it auto-exit if it stays idle for a long time and there's
no active inotify watches.

lei-daemon doesn't start until any other lei command is invoked;
so it shouldn't be started on installation.

> # show system-wide limits

> ==> /proc/sys/fs/inotify/max_user_instances <==
> 128

<snip>

> tail       367093 louis    4r  a_inode               0,14         0

Thats a lot of tail processes....
I wonder if they were spawned by `lei q -v' for emitting curl stderr?
They should be auto-killed.
(or if you have some other reason for running tail on your system).

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: Issues with `lei` as non-root
  2023-03-28  3:38         ` Eric Wong
@ 2023-03-28  4:08           ` Louis DeLosSantos
  2023-03-28 11:12             ` [PATCH] inotify: wrap with informative error message Eric Wong
  0 siblings, 1 reply; 9+ messages in thread
From: Louis DeLosSantos @ 2023-03-28  4:08 UTC (permalink / raw)
  To: Eric Wong; +Cc: meta

> Thats a lot of tail processes....
Ugh, sorry to waste your time.

All the tails were from a run away UI program I'm working on.
Once I killed them, `let` runs just fine as non-root.

Thanks for the free tech support, hope I didn't steal your attention
from something valuable :-D.

On Mon, Mar 27, 2023 at 11:38 PM Eric Wong <e@80x24.org> wrote:
>
> Louis DeLosSantos <louis.delos@gmail.com> wrote:
> > > Definitely not; the lei-daemon is per-user.
> >
> > Okay, maybe this is the issue to begin with? I installed lei from dnf.
> > I'm not sure what launches the daemon, is it launched on first run?
> >
> > If that is the case, it was probably launched when I restored to `sudo
> > lei q ....` command.
> > But, if its running as systemd service, I could move it to user service.
>
> You shouldn't need to manage it as a service; it's auto-started
> and killing it is harmless in most cases.  I'm considering it
> have it auto-exit if it stays idle for a long time and there's
> no active inotify watches.
>
> lei-daemon doesn't start until any other lei command is invoked;
> so it shouldn't be started on installation.
>
> > # show system-wide limits
>
> > ==> /proc/sys/fs/inotify/max_user_instances <==
> > 128
>
> <snip>
>
> > tail       367093 louis    4r  a_inode               0,14         0
>
> Thats a lot of tail processes....
> I wonder if they were spawned by `lei q -v' for emitting curl stderr?
> They should be auto-killed.
> (or if you have some other reason for running tail on your system).

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [PATCH] inotify: wrap with informative error message
  2023-03-28  4:08           ` Louis DeLosSantos
@ 2023-03-28 11:12             ` Eric Wong
  0 siblings, 0 replies; 9+ messages in thread
From: Eric Wong @ 2023-03-28 11:12 UTC (permalink / raw)
  To: Louis DeLosSantos; +Cc: meta

Louis DeLosSantos <louis.delos@gmail.com> wrote:
> > Thats a lot of tail processes....
> Ugh, sorry to waste your time.

Not at all.  Thank you for sharing your experience with us as it
will benefit future users, especially since lei is still in its
early days.  I think the patch below can help future users.

> All the tails were from a run away UI program I'm working on.
> Once I killed them, `let` runs just fine as non-root.

If you're adminning your own system (which it sounds like),
increasing the fs.inotify.max_user_instances should let you
run as many `tail' processes as you desire :>

Perhaps distros and/or kernel developers may raise limits
some day for single-user systems.

My distro uses /etc/sysctl.d/*.conf and /etc/sysctl.conf
to make those knobs persistent across reboots; but distros
vary/change a lot.

> Thanks for the free tech support, hope I didn't steal your attention
> from something valuable :-D.

No problem, and thank you for being one of the brave, early lei users :>
User feedback is valuable :>

------8<------
Subject: [PATCH] inotify: wrap with informative error message

As encountered by Louis DeLosSantos, Linux inotify is capped by
a lesser-known limit than the standard RLIMIT_NOFILE (`ulimit -n`)
value.  Give the user a hint about the fs.inotify.max_user_instances
sysctl knob on EMFILE, since EMFILE alone may mislead users into
thinking they've hit the (typically higher) RLIMIT_NOFILE limit.

I can test this on my system using:

  perl -I lib -MPublicInbox::Inotify -E \
   'my @x = map { PublicInbox::Inotify->new } (1..128)'

But I hesitate to include it in the test suite since triggering
the limit can cause unrelated processes to fail.

Link: https://public-inbox.org/meta/CAE6jdTo8iQfNM9Yuk0Dwi-ARMxmQxX-onL8buXcQ9Ze3r0hKrg@mail.gmail.com/
Reported-by: Louis DeLosSantos <louis.delos@gmail.com>
---
 MANIFEST                     |  1 +
 lib/PublicInbox/DirIdle.pm   |  4 ++--
 lib/PublicInbox/InboxIdle.pm |  4 ++--
 lib/PublicInbox/Inotify.pm   | 30 ++++++++++++++++++++++++++++++
 4 files changed, 35 insertions(+), 4 deletions(-)
 create mode 100644 lib/PublicInbox/Inotify.pm

diff --git a/MANIFEST b/MANIFEST
index 40535233..3c421645 100644
--- a/MANIFEST
+++ b/MANIFEST
@@ -217,6 +217,7 @@ lib/PublicInbox/In2Tie.pm
 lib/PublicInbox/Inbox.pm
 lib/PublicInbox/InboxIdle.pm
 lib/PublicInbox/InboxWritable.pm
+lib/PublicInbox/Inotify.pm
 lib/PublicInbox/InputPipe.pm
 lib/PublicInbox/Isearch.pm
 lib/PublicInbox/KQNotify.pm
diff --git a/lib/PublicInbox/DirIdle.pm b/lib/PublicInbox/DirIdle.pm
index 55c3982f..af99811c 100644
--- a/lib/PublicInbox/DirIdle.pm
+++ b/lib/PublicInbox/DirIdle.pm
@@ -9,14 +9,14 @@ use PublicInbox::Syscall qw(EPOLLIN);
 use PublicInbox::In2Tie;
 
 my ($MAIL_IN, $MAIL_GONE, $ino_cls);
-if ($^O eq 'linux' && eval { require Linux::Inotify2; 1 }) {
+if ($^O eq 'linux' && eval { require PublicInbox::Inotify; 1 }) {
 	$MAIL_IN = Linux::Inotify2::IN_MOVED_TO() |
 		Linux::Inotify2::IN_CREATE();
 	$MAIL_GONE = Linux::Inotify2::IN_DELETE() |
 			Linux::Inotify2::IN_DELETE_SELF() |
 			Linux::Inotify2::IN_MOVE_SELF() |
 			Linux::Inotify2::IN_MOVED_FROM();
-	$ino_cls = 'Linux::Inotify2';
+	$ino_cls = 'PublicInbox::Inotify';
 # Perl 5.22+ is needed for fileno(DIRHANDLE) support:
 } elsif ($^V ge v5.22 && eval { require PublicInbox::KQNotify }) {
 	$MAIL_IN = PublicInbox::KQNotify::MOVED_TO_OR_CREATE();
diff --git a/lib/PublicInbox/InboxIdle.pm b/lib/PublicInbox/InboxIdle.pm
index f0d8a972..4231c0a0 100644
--- a/lib/PublicInbox/InboxIdle.pm
+++ b/lib/PublicInbox/InboxIdle.pm
@@ -10,9 +10,9 @@ use parent qw(PublicInbox::DS);
 use PublicInbox::Syscall qw(EPOLLIN);
 my $IN_MODIFY = 0x02; # match Linux inotify
 my $ino_cls;
-if ($^O eq 'linux' && eval { require Linux::Inotify2; 1 }) {
+if ($^O eq 'linux' && eval { require PublicInbox::Inotify }) {
 	$IN_MODIFY = Linux::Inotify2::IN_MODIFY();
-	$ino_cls = 'Linux::Inotify2';
+	$ino_cls = 'PublicInbox::Inotify';
 } elsif (eval { require PublicInbox::KQNotify }) {
 	$IN_MODIFY = PublicInbox::KQNotify::NOTE_WRITE();
 	$ino_cls = 'PublicInbox::KQNotify';
diff --git a/lib/PublicInbox/Inotify.pm b/lib/PublicInbox/Inotify.pm
new file mode 100644
index 00000000..3ef271c8
--- /dev/null
+++ b/lib/PublicInbox/Inotify.pm
@@ -0,0 +1,30 @@
+# Copyright (C) all contributors <meta@public-inbox.org>
+# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
+
+# wrap Linux::Inotify2 XS module, support pure Perl via `syscall' someday
+package PublicInbox::Inotify;
+use v5.12;
+our @ISA;
+BEGIN {
+	eval { require Linux::Inotify2 };
+	if ($@) { # TODO: get rid of XS dependency
+		die "W: Linux::Inotify2 missing: $@\n";
+	} else {
+		push @ISA, 'Linux::Inotify2';
+	}
+};
+
+sub new {
+	$_[0]->SUPER::new // do {
+		my $msg = $!{EMFILE} ? <<EOM : "$_[0]->new: $!\n";
+inotify_init/inotify_init1: $!
+You may need to raise the `fs.inotify.max_user_instances' sysctl limit.
+Consult your OS documentation and/or sysctl(8) + sysctl.conf(5) manpages.
+EOM
+		$msg =~ s/^/E: /smg;
+		require Carp;
+		Carp::croak($msg);
+	}
+}
+
+1;

^ permalink raw reply related	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2023-03-28 11:12 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-03-28  1:00 Issues with `lei` as non-root Louis DeLosSantos
2023-03-28  1:32 ` Eric Wong
2023-03-28  1:36   ` public-inbox.org downtime expected in the next 6-8 hours Eric Wong
2023-03-28  2:30   ` Issues with `lei` as non-root Louis DeLosSantos
2023-03-28  2:52     ` Eric Wong
2023-03-28  3:05       ` Louis DeLosSantos
2023-03-28  3:38         ` Eric Wong
2023-03-28  4:08           ` Louis DeLosSantos
2023-03-28 11:12             ` [PATCH] inotify: wrap with informative error message Eric Wong

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).