[ClusterLabs Developers] Pacemaker
kgaillot at redhat.com
Mon Jan 9 10:43:06 EST 2023
On Mon, 2023-01-09 at 09:37 -0600, Ken Gaillot wrote:
> I do think the idea is good. I've commented on the pull request with
> some technical issues. It would be an API compatibility break, but
> coincidentally I'm planning such a break for a release at the end of
> this year, so that time frame could work.
Correction, the API break is planned for late next year (2024)
> What situation were you running into that this helped with? What kind
> of performance difference did you see in your testing?
> On Sat, 2023-01-07 at 23:44 +0300, Владимир Лямин wrote:
> > Could you write to me if I can continue to work in this direction?
> > сб, 7 янв. 2023 г. в 23:41, Владимир Лямин <
> > st067898 at student.spbu.ru>
> > :
> > > Hello, I'm Lyamin Vladimir. First-year master of St. Petersburg
> > > State University. I decided to optimize the running time of the
> > > pcmk__unpack_constraints function, since there is a loop over all
> > > the data here. I decided to use a hash table to optimize this.
> > >
> > > A hash table structure has been added, as well as functions to
> > > manage it.
> > > pe_resource_t* compareKey(const char* key, struct set *array);
> > > int getHash(const char *S);
> > > void push(Node **head, pe_resource_t* data);
> > > void insert(char* key, pe_resource_t* data, struct set *array);
> > > void init_array(struct set **array);
> > > void insert_children(pe_resource_t * rsc, struct set *hashTable);
> > >
> > > Existing functions have also been changed:
> > > pcmk__unpack_constraints
> > > (initialization of the hash table) and
> > > pcmk__find_constraint_resource (search for the desired resource)
Ken Gaillot <kgaillot at redhat.com>
More information about the Developers