Date: Tue, 13 Jan 1998 09:20:42 +0000 From: Edwin Rivera <edwinr@icanect.net> To: Dirk-Willem van Gulik <Dirk.vanGulik@jrc.it>, Doug White <dwhite@gdi.uoregon.edu> Cc: Dirk-Willem van Gulik <Dirk.vanGulik@jrc.it>, freebsd-questions@FreeBSD.ORG, MSQL/Perl Mailing list <msqlperl@franz.ww.tu-berlin.de> Subject: Re: mSQL-2.0.3 and FreeBSD 2.2.5 or 3.0 SNAP Message-ID: <3.0.32.19980113092038.009f0af0@icanect.net>
next in thread | raw e-mail | index | archive | help
Mr. van Gulik, Well, considering its infinite mode, where it went kernel-happy, it may be a cycling/resource limit issue. Perhaps splitting up the tasks into smaller chunks at a time may be a solution. Just my .02 Highest Regards, Edwin Rivera Webmaster / Programmer / Developer - ICANECT http://www.icanect.net ====================================================== At 09:10 AM 1/13/98 +0100, Dirk-Willem van Gulik wrote: >On Mon, 12 Jan 1998, Doug White wrote: > >> On Thu, 8 Jan 1998, Dirk-Willem van Gulik wrote: >> >> > On a WWW system; we are using the common combo >> > of Apache, Perl and mSQL. Since a recent change >> > where we have increased the free text and in >> > context search capabilities, we are regulary >> > seeing several of our index files of mSQL 2.0.3 >> > explode. Both on BSD 2.2.5 and 3.0 snap's. >> >> Oh joy. >> >> > We can reproduce it by cat-ting a 4 Mbyte >> > file of inserts into the 'msql DBASE' >> > command, so there seems little relation with >> > the perl/apache environment. >> > >> > The symptom is that a *.idx file grows quickly and >> > suddenly all time is spend in kernel land. With >> > no timely return; i.e. things like Alt-F1,2 etc >> > still work, as does pinging, telnet-connect; but >> > any typing or other user land feedback is >> > completely out. Another recent change is that we >> > now have some 150 tables, rather than the original >> > 25 or so. >> > >> > Anyone any ideas as where to look ? mmap() ? >> >> /etc/login.conf. I wonder if you are hitting the per-process limits. >> >> Try running `ulimit' or editing /etc/login.conf's limits. >> > >I set them to infinity... and this caused the machine to >go completely into kernel-land; with not a single slice >for user interaction... if I set it low (too low) I get >a decent error from the deamin that it has ran out of >disk-space, cycles, etc. So even though this might be >related; I cannot get any conclusive evidence. > >Dw. > > >---------------------------------------------------------------------- >To unsubscribe from this list send a message containing "unsubscribe" >to msqlperl-request@franz.ww.tu-berlin.de > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3.0.32.19980113092038.009f0af0>