100% CPU after 4.1.1 upgrade from 4.0.20

[henpa]

Hi, I've been running LiteSpeed for a long time and I don't remember the last time I had any problems with it.

Yesterday I upgraded from 4.0.20 to 4.1.1 (for no special reason) and later I started noticing an oddity: sometimes the litespeed process (lshttpd) used 100% of the CPU.

This happens when users use my service and when they try to upload a file, CPU of "litespeed" process goes to 100% and stays there....

I switched back and forth from 4.0.20 to 4.1.1 and I can reproduce this with 4.1.1 only.

I'm pretty sure it's a bug somewhere... but I haven't isolated the problem yet... this only happens when the user tries to upload a file (I tested with a 4MB file). I know I haven't done anything besides upgrading LiteSpeed.

This web service is a Perl script using CGI::Fast, on a Debian 5 linux running 2.6.26.

I did a "strace" to this process, and when the CPU goes to 100% it loops with messages like these:

10:45:23.817680 gettimeofday({1309959923, 817690}, NULL) = 0
10:45:23.817732 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.817785 gettimeofday({1309959923, 817795}, NULL) = 0
10:45:23.817836 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.817890 gettimeofday({1309959923, 817900}, NULL) = 0
10:45:23.817941 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.817994 gettimeofday({1309959923, 818004}, NULL) = 0
10:45:23.818046 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818099 gettimeofday({1309959923, 818109}, NULL) = 0
10:45:23.818150 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818204 gettimeofday({1309959923, 818214}, NULL) = 0
10:45:23.818255 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818308 gettimeofday({1309959923, 818318}, NULL) = 0
10:45:23.818360 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818413 gettimeofday({1309959923, 818423}, NULL) = 0
10:45:23.818464 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818518 gettimeofday({1309959923, 818528}, NULL) = 0
10:45:23.818569 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818622 gettimeofday({1309959923, 818632}, NULL) = 0
10:45:23.818674 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818727 gettimeofday({1309959923, 818737}, NULL) = 0
10:45:23.818778 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818832 gettimeofday({1309959923, 818841}, NULL) = 0
10:45:23.818883 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.818939 gettimeofday({1309959923, 818949}, NULL) = 0
10:45:23.818991 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1
10:45:23.819044 gettimeofday({1309959923, 819054}, NULL) = 0
10:45:23.819096 epoll_wait(6, {{EPOLLOUT, {u32=24, u64=24}}}, 10, 100) = 1

I switched back to 4.0.20. Let me know if I can help you with any other info!

Regards!

[cmanns]

time needed?

noatime enabled?

[henpa]

William, thanks for your reply, but I didn't understand what you mean. Could you please rewrite in "user-friendly language"?

Like I said, I only upgraded LiteSpeed from 4.0.20 to 4.1.1 from the admin web console. Is there an upgrade requirement that I am missing here?


Thanks

[mistwang]

Have you set "Inbound Bandwidth" limit?
http://www.litespeedtech.com/docs/we...y/#inBandwidth

Looks like a bug in bandwidth limit implementation.

[henpa]

hi mistwang, I never used this feature and I have both Outbound Bandwidth and Inbound Bandwidth set to 0.

Let me know if you need any other info.


Thanks

[mistwang]

Are you using a FastCGI script to handle the upload?

[henpa]

Yes I am! A Perl script with CGI::Fast module.

[mistwang]

can you please reproduce the bug with debug logging enabled?
You can search related log entries by your IP address, please send the log to bug@litespeedtech...

[henpa]

mistwang, I will do this tonight when users are not using my server ok? Thanks!

[mistwang]

sure. no problem.