Robert,
I agree with you.
Just explaining how it is right now. The best we can do right now is to
document which parameters require --initial.
You can always do a --initial restart, it takes a bit longer though...
BR,
T
Casto, Robert wrote:
>Tomas,
>
>How are we supposed to know if a change will affect the file system? My guess would be
> that all changes would. How about making --initial safer or automatic? Then we would not
> have to worry about using it or not.
>
>Robert Casto
>Quality Measurements
>206.266.3695
>
>
>
>-----Original Message-----
>From: Tomas Ulin [mailto:tomas@stripped]
>Sent: Wednesday, January 12, 2005 1:22 PM
>To: jon stuart
>Cc: cluster@stripped
>Subject: Re: resizing ndb storage nodes
>
>Jon,
>
>jon stuart wrote:
>
>
>
>>hi there,
>>
>>when i'm increasing DataMemory, IndexMemory and the like, i understand
>>the procedure to be
>>
>>1. edit ndb_mgmd's config.ini
>>2. restart ndb_mgmd
>>3. restart each ndbd in turn,
>> waiting for it to finish starting before moving to the next.
>>
>>
>>
>correct
>
>
>
>>assuming this is correct, my question is: when restarting the ndbd's,
>>is it enough to do perform
>>
>> ndb_mgm> 1 restart
>>
>>
>>
>yes (but in most cases, see below)
>
>
>
>>or do i need to pass the --initial flag to a manual invocation of ndbd
>>on each node?
>>
>>
>>
>ndb_mgm> 1 restart -i (ndb_mgm equivalent to --initial)
>
>must be performed if the changes affects the filesystem, e.g.
>NoOfFragmentLogFiles
>
>
>
>
>
>
>>from looking at the source for ndb_mgmd, it looks like the -i flag to
>>restart might be equivalent to passing --initial, but i'd like to be
>>sure.
>>
>>
>>
>correct
>
>
>
>>cheers, jon.
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>
>
