You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The web UI seems to use different methods to figure out table/replica availability in different places. This sometimes leads to inconsistent data (see the attached image).
Also I often noticed that on the overview for a specific table, right under Statistics, it would say "Live" next to a green icon, even if some of the shards are actually down. In this example, the table was in fact not available, neither for outdated nor up-to-date reads.
Hum, looking at the code, we just look at the blueprint and check if there is enough available replicas.
By available replica, we mean a replica that is on a machine that is not down.
I'll update this code to use the same as on the Replication Status view (checking the blueprint and directory).
Up in review 1189 assigned to @mglukhovsky
Branch michel_1756_live_status
A table is "Live" only if every shards have enough replicas up to date to meet the number of acks required, which mean a table can be "Live" even if it's replicating.
The number of master/replicas reachable in the profile view (first section) is the number of shards/replicas own by a machine that we can reach.
I renamed shards/replicas reachable in the sections below to shards/replicas available since in this case we check if we can use them or not.
I also fixed a bug for unsatisfiable goal in my commit in case universe is master (we forgot to add 1 in this case).
The web UI seems to use different methods to figure out table/replica availability in different places. This sometimes leads to inconsistent data (see the attached image).
Also I often noticed that on the overview for a specific table, right under Statistics, it would say "Live" next to a green icon, even if some of the shards are actually down. In this example, the table was in fact not available, neither for outdated nor up-to-date reads.
@neumino: Ideas? :-)
The text was updated successfully, but these errors were encountered: