In article <8a3hnf$h82$1 at icrf.news>, <nospam at our.site> wrote:
>This message has been posted by: Arne Mueller <a.mueller at REMOVE-THIS-TO-SENDicrf.icnet.uk>
>>Gary Williams wrote:
>>>> We are seeing increasing numbers of BLAST searches failing like this:
>>I had problems with blast version 2.0.6 (I guess), sometimes it crashed
>and it took me while to look at the syslog of our server to find out
>that it was killed by the operating system (IRIX 6.5).
>>The reason was that blastpgp installed a signal handler to catch
>segmentation faults and this handler itself caused a segmentation fault.
>To prevent an endless recoursion it was killed by the os (that was the
>message in the syslog). It's along time ago and I don't realy remember
>;-( but I think blastpgp was terminated without printing any message.
>>Is there anything in your syslog?
Our sysadmin (Peter Tribble) says:
"Nope. I don't think system resources have anything to do with it.
It crashes the same on Sparc and Intel, and it doesn't seem to depend
on the version or whether you use the precompiled executables or build
your own (although I suspect that, as we use the same compilers as
NCBI do, there's not actually any difference).
Getting NCBI blast to actually run reliably seems rather difficult.
Most of the benchmarks I've run have crashed out"
Gary Williams Tel: +44 1223 494522 Fax: +44 1223 494512
mailto:G.Williams at hgmp.mrc.ac.ukhttp://www.hgmp.mrc.ac.uk/
Bioinformatics,MRC HGMP Resource Centre,Hinxton,Cambridge, CB10 1SB,UK