While load testing an xcache it was noticed that after the end of a period of heavy load there were a number of files left open on the server (several minutes after all clients had terminated). There were also a number of threads left on the server in this position:

Thread 1853 (Thread 0x7fb6ee7ff700 (LWP 278761)):
#0  0x00007fb73d3a8b3b in do_futex_wait.constprop.1 () from /lib64/libpthread.so.0
#1  0x00007fb73d3a8bcf in __new_sem_wait_slow.constprop.0 () from /lib64/libpthread.so.0
#2  0x00007fb73d3a8c6b in sem_wait@@GLIBC_2.2.5 () from /lib64/libpthread.so.0
#3  0x00007fb73d830596 in XrdLink::Serialize() () from /lib64/libXrdUtils.so.3
#4  0x00007fb73dad65d8 in XrdXrootdProtocol::do_Close() () from /lib64/libXrdServer.so.3
#5  0x00007fb73d833ec9 in XrdLinkXeq::DoIt() () from /lib64/libXrdUtils.so.3
#6  0x00007fb73d83746f in XrdScheduler::Run() () from /lib64/libXrdUtils.so.3
#7  0x00007fb73d8375b9 in XrdStartWorking(void*) () from /lib64/libXrdUtils.so.3
#8  0x00007fb73d7f6a57 in XrdSysThread_Xeq () from /lib64/libXrdUtils.so.3
#9  0x00007fb73d3a2ea5 in start_thread () from /lib64/libpthread.so.0
#10 0x00007fb73c6a39fd in clone () from /lib64/libc.so.6

The clients which were generating the test load had logged a number of failures of XrdCl::File::Close reported as "[ERROR] Socket error". In the cases examined the client had received all the requested data correctly. Some investigation pointed to a failure of XrdXrootdAioReq::Read() on the first read() of a request, due to no available XrdXrootdAio objects. I think this is due to XrdXrootdAioReq::Alloc() sometimes returning a request object without the minimum number of XrdXrootdAio being available. I'll open a suggested pull request in a moment.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or unsubscribe.

[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/xrootd/xrootd/issues/1407", "url": "https://github.com/xrootd/xrootd/issues/1407", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

Use REPLY-ALL to reply to list

To unsubscribe from the XROOTD-DEV list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1