Forum Discussion

BaltoStar_12467's avatar
Aug 14, 2014

BIG-IP : iControl LocalLBDataGroupFile.set_local_path()

F5 BIG-IP LTM VE v11.4.0 on ESXi

iControl re-cache data-group operation :


For a live prod BIG-IP cluster with a VIP iRule that is actively reading data from the data-group , is the above considered a risky operation ?

e.g. attempt live-swap of data-group's data-file, fails due to locks or other systems-level issues, data-group now unavailable to iRule

I've attempted live-update of a data-file's contents via BIG-IP browser admin ( copy/paste new contents and click “Update” ) and seen the corresponding data-group become unavailable to the iRule. The paranoid approach is to create a new data-group/file and a copy of the iRule modified to point to this new data-group , and then in the VIP’s iRule list swap old/new iRules – so that never actually update a “live” data-group.

So now I am concerned about performing similar operation via iControl

–- although apparently this API is not actually live-updating a cached data-file but rather re-pointing data-group to a new cached data-file ?

2 Replies

  • "...seen the corresponding data-group become unavailable to the iRule."


    Really? Thats very surprising. Adding entries to a datagroup from the GUI should never take it offline.


    "eg. attempt live-swap of data-group's data-file..."


    Why would you need to do this? The F5 has a running config and a stored config. When you update a datagroup which points to a file you only need to update the file, then perform a config load to have that new files contents loaded into memory. It is also indexed at this time.


  • Thanks for the response Kevin.

    Please note that I am on BIG-IP v11. The data-group mechanisms you describe I believe apply to pre-v11.

    New to v11, the operation

    immediately reloads & recaches an external data-group with the contents of the supplied data-file -- there's no need to trigger BIG-IP to reload its config.

    It is not recommended to directly update the local data-files BIG-IP self-maintains for each external data-group ( if you could even locate the file currently being used ). On create/update of an external data-group, BIG-IP auto-creates a new persistent data-file stored in a special directory. This data-file's name is constructed as the data-group name appended with a unique identifier and revision index to track history of changes.

    My concerns are that on reload/recache of very large data-sets ( 1M+ rows ) something goes wrong and the data-group becomes unavailable to those iRules that reference it.