[Not Resolved] LSWS 4.1 - Spinlock

[JLHC]

Originally Posted by mistwang

We fixed a few bugs that cause 4.1 crash, however, for this particular one, we still need a more detail report running LSWS under valgrind.

Any update to this yet? Planning to upgrade to version 4.1.x but this issue is still holding us back for now.

[JLHC]

I can see that version 4.1.1 has been released. Is this solved in the new version?

[JLHC]

I can see that version 4.1.2 has been released. Is this solved in the new version?

[MikeDVB]

I haven't run into the issue in a while, but the issue only occurred when the maximum connections was set to something around 1k~9k. We have it set to something like 50k on each server which has totally prevented the issue from causing issues, not necessarily from occurring in the first place.

The only way for me to know for sure is to drop the servers back to a limit of 5k simultaneous connections and then wait a few days.

[JLHC]

Originally Posted by MikeDVB

I haven't run into the issue in a while, but the issue only occurred when the maximum connections was set to something around 1k~9k. We have it set to something like 50k on each server which has totally prevented the issue from causing issues, not necessarily from occurring in the first place.

The only way for me to know for sure is to drop the servers back to a limit of 5k simultaneous connections and then wait a few days.

Well that will be troublesome for you.

I hope LiteSpeed Tech's staffs can chime in to clarify this. They didn't say anything about this in the Version History / Release Log.

[JLHC]

I can see that version 4.1.3 has been released. Is this solved in the new version?

[bbb177]

The issue is still present. I'm using the newest version and it "hangs" at least once or twice a day. Are you at least working on a solution? I'm paying for a server that doesn't work like it should. And going back to 4.020 is not an option because I need the mp4 streaming. That's the main reason that led me to buying a license. I really need to know when the issue will be resolved.

[NiteWave]

we'd like to solve such issues as soon as possible, the problem is it's hard to reproduce.

have you tried MikeDVB's suggestion:

I haven't run into the issue in a while, but the issue only occurred when the maximum connections was set to something around 1k~9k. We have it set to something like 50k on each server which has totally prevented the issue from causing issues, not necessarily from occurring in the first place.

[bbb177]

Originally Posted by NiteWave

we'd like to solve such issues as soon as possible, the problem is it's hard to reproduce.

have you tried MikeDVB's suggestion:

Yes, and it doesn't work, the server still "hangs". As far as I know my hosting sent bug reports to you, so you should have detailed knowledge how to reproduce it.

[jackc]

I'm seeing the same problem too, with the latest 4.1.10, didn't have this problem before with heavier server load and older version of lsws, really weird.
when the problem occured litespeed is using 100 percent cpu. also seeing possibly memory leak, litespeed is using alot more memory than usual over time.

some strace below

epoll_ctl(7, EPOLL_CTL_DEL, 59, {0, {u32=59, u64=59}}) = 0
close(59) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 78, {0, {u32=78, u64=78}}) = 0
close(78) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 115, {0, {u32=115, u64=115}}) = 0
close(115) = 0
recvfrom(136, 0x7fff2ba1daef, 1, 2, 0, 0) = -1 EAGAIN (Resource temporarily unavailable)
epoll_ctl(7, EPOLL_CTL_DEL, 174, {0, {u32=174, u64=174}}) = 0
close(174) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 175, {0, {u32=175, u64=175}}) = 0
close(175) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 256, {0, {u32=256, u64=256}}) = 0
close(256) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 265, {0, {u32=265, u64=265}}) = 0
close(265) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 266, {0, {u32=266, u64=266}}) = 0
close(266) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 268, {0, {u32=268, u64=268}}) = 0
close(268) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 272, {0, {u32=272, u64=272}}) = 0
close(272) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 278, {0, {u32=278, u64=278}}) = 0
close(278) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 279, {0, {u32=279, u64=279}}) = 0
close(279) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 288, {0, {u32=288, u64=288}}) = 0
close(288) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 393, {0, {u32=393, u64=393}}) = 0
close(393) = 0
epoll_ctl(7, EPOLL_CTL_DEL, 425, {0, {u32=425, u64=425}}) = 0
close(425) = 0