On 02/04/2013 12:16 PM, Rich Megginson wrote:
On 02/04/2013 09:42 AM, Mark Reynolds wrote:
There has been a request to remove instance specific scripts, and
make them generic and placed in /usr/sbin/
https://fedorahosted.org/389/ticket/528
Should we remove all the scripts from /usr/lib64/dirsrv/slapd-INSTANCE:
bak2db db2index dn2rdn ldif2ldap
ns-newpwpolicy.pl start-slapd usn-tombstone-cleanup.pl
bak2db.pl db2index.pl fixup-linkedattrs.pl
monitor restart-slapd stop-slapd verify-db.pl
cleanallruv.pl db2ldif fixup-memberof.pl ns-accountstatus.pl
restoreconfig suffix2instance vlvindex
db2bak db2ldif.pl ldif2db ns-activate.pl
saveconfig syntax-validate.pl
db2bak.pl dbverify ldif2db.pl ns-inactivate.pl
schema-reload.pl upgradednformat
It appears almost of these are all instance specific. Do we want to
make all of these "generic" and stick them in /usr/sbin?
Yes.
Works for me.
However, I think it makes sense to keep some instance specific
scripts: monitor, stop/start/restart, maybe the "task" scripts too
(like schema reload, cleanallruv, fix-memberof, etc). Then only make
the database scripts generic: db2ldif db2index, ldif2db, db2bak,
bak2db, etc.
One of the goals is to get any "dynamic" files out of /usr. Most
installations want to mount /usr read-only. They do not want to mount
/usr read-write in order to create or update a 389 instance.
IPA works around this by putting the instance specific scripts in
/var/lib/dirsrv/scripts-INSTANCE - but most people aren't looking for
389 maintenance commands under /var.
I don't think we need any instance specific scripts, but if we do,
they should not be in /usr or /var.
Ok, I don't think we need them, but if someone really wants them, they
could go in a new directory under /etc/dirsrv/*
We have already started to do this with the
start-dirsrv/stop-dirsrv/restart-dirsrv scripts in /usr/sbin. With no
argument, they work on all instances, or accept the instance name as
an argument. This is fine for start/stop/restart, but might be a bit
surprising for something like db2ldif, if it then does an export of
all instances. For all scripts other than start/stop/restart, I think
the command should fail if the instance is not specified e.g.
# db2ldif ...
Error: More than one instance - specify the instance name as one of:
foo bar ...other instance names...
One request was to check if there is only a single instance and then use
that one by default, otherwise error out. I think this makes sense, and
should be easy to do.
The other nice thing about the start-dirsrv etc. scripts is that they
use the instance information from /etc/sysconfig/dirsrv-* to look for
the dse.ldif, log files, etc. That removes the need for an instance
specific script created during instance creation time with "hardcoded"
instance specific values.
Yup, sounds like a plan.
Thanks for your input Rich!
Mark
Just curious what everyone thinks about this.
Thanks,
Mark
--
Mark Reynolds
Red Hat, Inc
mreynolds@xxxxxxxxxx
--
389-devel mailing list
389-devel@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/389-devel
