I've been working to try and release Samba4 TP3, and I've been testing
SWAT, as we don't currently have an automated test for that part of the
platform.
I have already fixed some issues, but I'm a little perplexed as to how
to debug the provision functionality: When I provision, it runs for a
while, then just displays a blank page to the browser
Andrew Bartlett

Tue, 2006-10-10 at 15:30 +1000, Andrew Bartlett wrote:
I've been working to try and release Samba4 TP3, and I've been testing
SWAT, as we don't currently have an automated test for that part of the
platform.

I have already fixed some issues, but I'm a little perplexed as to how
to debug the provision functionality: When I provision, it runs for a
while, then just displays a blank page to the browser

This is currently blocking the Samba4 TP3 release.

As far as I can tell, upping the timeout doesn't help. Neither does
removing the timeout from the actual running of the EJS engine.

Andrew Bartlett

Thu, 2006-10-12 at 18:43 +1000, tridge (AT) samba (DOT) org wrote:
Andrew,

This is currently blocking the Samba4 TP3 release.

I've fixed the provisioning speed problem.

Main problem was an interaction between provisioning and existing
partitions with a partitions module. The provisioning code tries to
wipe the ldb in a 'friendly' way in case processes are already
attached to it, by searching for all records and deleting them. This
wasn't working with partitions, as searches don't cross partition
boundaries. That caused a second provision after the system was
already installed to proceed very slowly, as it had to do a lot of
work re-indexing the (large!) schema on each record.

I'm a bit confused. As the code stood, it did delete the data in the
partitions. If it did not, then loading the new records would fail.

Furthermore, the new code deletes the contents of the 'old' partitions,
but I'm worried: If the new provision is configured not to use LDAP
(for example), we will now go off and wipe the old LDAP server.

That is why the old code wiped man DB, initialised the partitions
records, and then wiped the partitions data.

Andrew Bartlett

Thu, 2006-10-12 at 21:18 +1000, Andrew Bartlett wrote:
Thu, 2006-10-12 at 18:43 +1000, tridge (AT) samba (DOT) org wrote:
Andrew,

This is currently blocking the Samba4 TP3 release.

I've fixed the provisioning speed problem.

Main problem was an interaction between provisioning and existing
partitions with a partitions module. The provisioning code tries to
wipe the ldb in a 'friendly' way in case processes are already
attached to it, by searching for all records and deleting them. This
wasn't working with partitions, as searches don't cross partition
boundaries. That caused a second provision after the system was
already installed to proceed very slowly, as it had to do a lot of
work re-indexing the (large!) schema on each record.

I'm a bit confused. As the code stood, it did delete the data in the
partitions. If it did not, then loading the new records would fail.

Furthermore, the new code deletes the contents of the 'old' partitions,
but I'm worried: If the new provision is configured not to use LDAP
(for example), we will now go off and wipe the old LDAP server.

That is why the old code wiped man DB, initialised the partitions
records, and then wiped the partitions data.

I've changed the code again, to address this. The main issue is that
the erasure of the partitions occurred *after* the index was added.
I've pushed the index creation to the very end of the provision.

Andrew Bartlett