[Debian] AW: Bug#557785: vzctl: symlinked config file: symlink overwrittenwhen --save isspecified

[Robert Heinzmann]

Olá Ola :)

Is this a data corruption issue ? I think not in the first place, but It can become one.

Consider the typical HA scenario (* is the active config and node):

[NODEA*]-- --- [NODEB]
          |
        [CFG*]
   (shared storage 
       / drbd)

Let's assume you are changing the IP of the machine, or some other config parameter like some disk device on NODEA with --save. What happens is: 


Before Failover, after --save: 

[NODEA*]    --- [NODEB]
  |        |
[modCFG*] [CFG]


Everythign runs fine. Now if you are NOT aware of the bug, if you failover, your old config is restored and you may access wrong devices, use a wrong IP (duplicate?) etc ... This probably happens weeks after the change (--save) and you will have trouble finding the issue.

After Failover: 

[NODEA]    --- [NODEB*]
  |        |
[modCFG] [CFG*]

So for me it seems important that this bug can be found by people looking for this. If it needs fixing - yes, but how ? 

A) 

* Fix in stable 
(This will probably not break any existing configurations, it will just keep links that were removed before)

Pro: would help fast
Pro: no side effetcs
Con: does it break anything ? 

B) 

* Mark this bug as beeing a known bug up and including 3.0.23 in lenny (so that other peaople find it)
* Fix in 3.0.23 (patch) 
* Backport 3.0.23 to lenny

Pro: can be done now
Con: what is 3.0.23 - is it a stable release ? 
Con: maybe side effects with 3.0.23


C) 

* Mark this bug as beeing a known bug up and including 3.0.23 in lenny (so that other peaople find it)
* Wait for 3.0.24 in sid (in the meantime live with a wrapper script/job)
* Backport 3.0.24 to lenny

Pro: 3.0.24 will be the next stable
Con: When will 3.0.24 arrive ?  


I think B) is not the best option. For me A) would be nice, hoever I can also live with C) for the sake of stability.

Regards,
Robert