fix/Estimations announcement #2597
Conversation
Codecov Report
@@ Coverage Diff @@
## master #2597 +/- ##
=======================================
Coverage 29.71% 29.71%
=======================================
Files 407 407
Lines 31143 31145 +2
=======================================
+ Hits 9254 9255 +1
Misses 21078 21078
- Partials 811 812 +1
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
Seems like we also need a routine that is going to iterate over containers and be ready to clean up every info on every container removal (also, the storage engine should support this operation too, of course). But it is a too big change so not added it yet to this PR. We have some removal subscription but it won't work for the removals at the SN's offline time. Thoughts? @roman-khimov, @cthulhu-rider, @AliceInHunterland |
carpawell
force-pushed
the
fix/estimations-announcement
branch
2 times, most recently
from
afa9110
to
053137b
Compare
Are you talking about #1663 or some kind of locally stored metabase thing related to estimations specifically? |
Like this, yes. If a node is sleeping it won't be able to drop its indexes with the notifications approach (that we already have). But yes, now it does absolutely nothing if we are talking about the estimations but I do not want to fix it fast cause notifications is only a part of potentially useless info in the metabase if a container is dropped, so want to hear your thoughts and maybe create another issue for more accurate solution. But the current missing container issue is fixed with this PR, although I would not treat it as the best solution for the exact that problem. |
carpawell
force-pushed
the
fix/estimations-announcement
branch
from
053137b
to
493f081
Compare
|
I'd forget this problem for now, estimation mechanism itself won't be like this forever. |
carpawell
force-pushed
the
fix/estimations-announcement
branch
from
493f081
to
224cf70
Compare
|
Needs a rebase, CHANGELOG has changed. |
Iterations over the locally calculated values should not stop because of any errors. Sending part of the information is better than sending nothing in this case. Moreover, the `code = 3072 message = container not found` logs were found, they are definitely not a good reasons to stop the iterations over the remaining estimations. Refs nspcc-dev#1506. Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
In particular, it would save from estimating removed containers found in the logs (they were empty), but it is also useless in general. Refs nspcc-dev#1506. Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
`ListContainers` do not use a string container ID representation so do not do that useless conversion. Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
carpawell
force-pushed
the
fix/estimations-announcement
branch
from
224cf70
to
3e4e712
Compare
No description provided.