You are looking at the compat_mode branch. Old versions of the code
could indeed only backfill one osd at a time. I suggest you trace
back that variable to understand when it's set.
-Sam
On Mon, Nov 14, 2016 at 11:21 PM, xxhdx1985126 <xxhdx1985126@xxxxxxx> wrote:
>
> Hi, everyone.
>
>
> When I read the source code of PG::calc_replicated_acting method, I found the following code really confusing:
>
>
>
> // select replicas that have log contiguity with primary.
> // prefer up, then acting, then any peer_info osds
> for (vector<int>::const_iterator i = up.begin(); i != up.end(); ++i) {
> pg_shard_t up_cand = pg_shard_t(*i, shard_id_t::NO_SHARD);
> if (up_cand == primary->first)
> continue;
> const pg_info_t &cur_info = all_info.find(up_cand)->second;
> if (cur_info.is_incomplete() ||
> cur_info.last_update < MIN(
> primary->second.log_tail,
> auth_log_shard->second.log_tail)) {
> /* We include auth_log_shard->second.log_tail because in GetLog,
> * we will request logs back to the min last_update over our
> * acting_backfill set, which will result in our log being extended
> * as far backwards as necessary to pick up any peers which can
> * be log recovered by auth_log_shard's log */
> ss << " shard " << up_cand << " (up) backfill " << cur_info
> << std::endl;
> if (compat_mode) {
> if (backfill->empty()) {
> backfill->insert(up_cand);
> want->push_back(*i);
> acting_backfill->insert(up_cand);
> }
> } else {
> backfill->insert(up_cand);
> acting_backfill->insert(up_cand);
> }
> } else {
> want->push_back(*i);
> acting_backfill->insert(up_cand);
> usable++;
> ss << " osd." << *i << " (up) accepted " << cur_info << std::endl;
> }
> }
>
>
> It seems that only when the "backfill" set is empty will an OSD in up set whose pg log can not be linked up with the pg log of auth_log_shard or primary be added into "backfill" set, which means only one OSD can be "backfilled".
> What if there are two or more OSD in up set that needs to be "backfilled"?
>
>
> Please help me. Thank you.
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
