This wiki is obsolete, see the NorduGrid web pages for up to date information.

ARC 0.8.1b TODO

From NorduGrid
Jump to navigationJump to search

ARC 0.8.1b is a pre-release of 0.8.1

Availability

The 0.8.1b packages is available from download.nordugrid.org through direct downloads or via the yum/apt "testing" repositories.

Known issues with 0.8.1b

Problem Status
grid-infosys fails to start on some systems Fixed in RC2
BDIIv4 problems on new systems Not solved
LRMS Queues complain about expired host certificate Fixed in RC3 or final
String not supported in walltime Fixed in RC3 or final
Problems downloading CGI scripts with arguments Fixed in RC3 or final

Information system

grid-infosys fails to start on some systems

  • Problem with newer versions of autotools (2.60) resulted in a bad grid-infosys startup script. (See: Bug #1635))
  • Runtime fix: remove the string: "../../" in /etc/init.d/grid-infosys

BDIIv4 problems on new systems

The current BDIIv4 is known to fail on systems with slapd linked against libdb-4.7.so. The symtoms are that the infosystem works briefly and then queries will either be: veryslow, hang or fail. Kernel messages like the following will appear:

 Oct 16 11:10:45 hostname kernel: bdii4-fwd[28985]: segfault at 65c ip 005c4919 sp bf928a40 error 4 in libperl.so[518000+256000]

The bug related to this is: Bug #1617)

LRMS Queues complain about expired host certificate

  • Symptom: ngsub fails to submit jobs to the cluster and complains about expired host certificate.
  • Solution: Add something like the following to the [common] or [infosys] section of /etc/arc.conf:
 x509_cert_dir="/etc/grid-security/certificates"

This should be default and will be fixed in the final 0.8.1 release.

  • Status: Fixed in 15241 and 15243

String not supported in walltime

With xRSL entry:

 (walltime="3 hours")

you get:

 Job submission failed due to: Can not convert string to int: 3hours

The problem is reported in: Bug #1649

  • Temporary solution: Express the time in minutes
  • Fixed in revision 15235

Problems downloading CGI scripts with arguments

Downloading urls such as:

 http://example.com/script.php?attr=val

with eg. ngcp in 0.8.1rc1 fails. This works with 0.8.0 and arccp. Tracked in: Bug #1686

Status: Fixed in revision 15229 and 15230

ARC Information Index Server (EGIIS)

Due to the way that the new ARC Information Index Server is implemented it is not supported by all platforms. It is known not to work with out-of-the box openldap on the following systems:

  • Gentoo (slapd not a shared object)

Distributions which have been checked and should work:

  • Red Hat Enterprise 5 i386,x86_64
  • Red Hat Enterprise 4 i386
  • CentOS 5 i386,x86_64
  • Fedora 11 i386, x86_64

Post RC2, these should work as well:

  • Debian 5.0.x AMD64
  • Ubuntu 8.04 i386
  • Ubuntu 8.10 i386
  • Ubuntu 9.04 x86_64

The reason is that EGIIS uses the standard openldap server, but modifies the functionality slightly so it is backwards compatible with old ARC. Globus implemented a non-standard LDAP server (MDS GIIS) which responds with a whole LDAP entry even though the single attribute (giisregistrationstatus) is requested.