Project

General

Profile

Feature #2685

Factory reconfig could detect and fail on missing entries

Added by Burt Holzman over 8 years ago. Updated over 8 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Douglas Strain
Category:
-
Target version:
Start date:
04/27/2012
Due date:
% Done:

0%

Estimated time:
Stakeholders:
Duration:

Description

We've known about this problem for a while -- you can't delete entries. (Well, you can, but you can't re-add them).

Here's Jeff's mail on it:

??We encountered a condition that can lead to problems during reconfig. It seems the factory allows you to reconfig with a glideinWMS.xml file that has a missing entry in it. However after this is done if we discover it and put the entry back in the config and try to configure again it fails. Here's how it happened at cern:

1. Marian copied glidein_v2_1.cfg/glideinWMS.xml to glidein_v2_1.cfg/glideinWMS.120425.xml then added a new entry to glideinWMS.120425.xml and used that one to reconfig.
2. I didn't know this and I made updates to glidein_v2_1.cfg/glideinWMS.xml and used that one to reconfig. The new entry marian created got lost and no warning messages.
3. Marian realized the new entry was missing and put it back. This time when he ran reconfig:
vocms107] /home/gfactory/glideinsubmit/glidein_v2_1 > ./factory_startup
reconfig ../glidein_v2_1.cfg/glideinWMS.120427.xml
reconfig_glidein { -force_name name -writeback yes|no -update_scripts
yes|no -xml xml | -help }

Cannot create submit dir
/data/srv/gfactory/config/glideinsubmit/glidein_v2_1/entry_CMS_T1_DE_KIT_HTPC_cream-1,
already exists.
Reconfiguring the factory [FAILED]

I suggest what should happen is the reconfig I did on 2 should detect an entry was missing and bail out with an error message. This way I would have realized I was editing a stale config file and no harm would have been done.??


Related issues

Precedes GlideinWMS - Feature #2693: Add --force-delete-entry option to factory reconfigClosed05/02/2012

History

#1 Updated by Burt Holzman over 8 years ago

  • Status changed from New to Assigned

We just discussed this over lunch with Jeff Dost.
For starters, it's sufficient for a reconfig to error out if entries are deleted and complain vociferously.
(Bonus points if we point to the "live" glideinWMS.xml that has the old entry in it, so an operator can
recover from accidentally removing entries without backing up their own config file.

It may be useful to have an option to force deleting entries as well, but that's under discussion. I'll add a low-priority subtask for it.

#2 Updated by Douglas Strain over 8 years ago

  • Status changed from Assigned to Resolved

This has been done in branch_v2plus_2685 and branch_master_2685.

The following changes take effect:
- Deletion of entries will cause an abort of the reconfig
- Addition of the "-force_delete" parameter to factory_startup to override
- Re-creation of old entries does not cause an error
Note: I haven't actually changed the deletion behavior, so it keeps the
old directories etc there.

Commits:
v2plus: bbb87bf9cb2f571810516d337bf81780f4d65c9a
master: 4eba71d55c2737e3b6ee6254bf53458e3c9e3358 , c3efe11bad3d6ee37a69eb04f24abf69caf82675 (template change specific to v3+)

#3 Updated by Parag Mhashilkar over 8 years ago

Douglas Strain wrote:

This has been done in branch_v2plus_2685 and branch_master_2685.

The following changes take effect:
- Deletion of entries will cause an abort of the reconfig
- Addition of the "-force_delete" parameter to factory_startup to override
- Re-creation of old entries does not cause an error

Note: I haven't actually changed the deletion behavior, so it keeps the
old directories etc there.

Commits:
v2plus: commit:bbb87bf9cb2f571810516d337bf81780f4d65c9a
master: commit:4eba71d55c2737e3b6ee6254bf53458e3c9e3358 , commit:c3efe11bad3d6ee37a69eb04f24abf69caf82675 (template change specific to v3+)

#4 Updated by Parag Mhashilkar over 8 years ago

  • Status changed from Resolved to Assigned

Reviewed. Should be good to merge.

#5 Updated by Parag Mhashilkar over 8 years ago

  • Status changed from Assigned to Resolved
  • Target version changed from v2_7_x to v2_6

#6 Updated by Parag Mhashilkar over 8 years ago

  • Status changed from Resolved to Closed

Also available in: Atom PDF