Skip to content

Merge pull request #10861 from apple/ahusain/releaseNotes-71.3.3 - #7

Open
mo74m3ed wants to merge 521 commits into
mainfrom
snowflake/release-71.3
Open

Merge pull request #10861 from apple/ahusain/releaseNotes-71.3.3#7
mo74m3ed wants to merge 521 commits into
mainfrom
snowflake/release-71.3

Conversation

@mo74m3ed

Copy link
Copy Markdown
Owner

PLEASE DELETE THIS TEMPLATE WHEN YOU POST YOUR PR FOR REVIEW

Please describe:

  • The problem (for bugs: how/where you encountered it).
  • The solution.
  • Testing done, if any.

We encourage AI-assisted review. Save your diff:

git diff upstream/main -- . > /tmp/pr.diff

Then review it with a prompt like:

Review /tmp/pr.diff. What is it trying to do? Is it correct? Are there bugs? Are there omissions? Are there better ways of doing things? Should this CL be LGTMd?

END OF TEMPLATE TO DELETE

sfc-gh-satherton and others added 30 commits April 26, 2023 10:40
…t in most workloads that create metaclusters
apple#10068)

* make status json report desired role count (log, resolver, proxies...) regardless whether user sets it explicitly

* make proxies calculation simpler

* Add comment for DatabaseConfiguration
…(Cherry-pick apple#10041 to snowflake/release-71.3)) (apple#10042)

* Fix for correctness failures when issuing duplicate requests
Add comments; Disable failure injection in Snap test

* Add Snapshot tests which are clean when fault injection is disabled in Snaptest.

* Remove snapshot restarting tests in from_7.1.0_until_71.2.0
…ot test failure. (apple#10078)

* Downgrade "inaccurate shard estimate" consistency check to warning, not test failure.

We know that the hash function used for sampling keys is biased and oversamples.
The fix is non-trivial, since we can't straightforwardly change the sampling function
for existing FDB databases. Thus, we downgrade this particular test failure to a warning.

* Disable test failure entirely, since it will retry and fail again.

Change the existing log message to a warning, so we at least have
some indiciation in the logs that we hit the biased shard estimate
issue.
…idate it when loading the entry from the cluster. (apple#10085)
…#10073 to snowflake/release-71.3) (apple#10099)

* Write storage_metadata for TSS and add test for it (Cherry-Pick apple#10073 to snowflake/release-71.3)

* populate storage metadata for tss

* add metadata check to quiesent consistency check

* check storage metadata and storage server interface in the same transaction
…we have Python 2 files and a few files written externally. Add external files as exclusions to the precommit checks. (apple#10094)
…tdown improvements found along the way. (Cherry-Pick apple#10109 to snowflake/release-71.3) (apple#10131)
…10108) (apple#10115)

Description

'Type' is a reserved keywork for TraceEvent, RESTKmsConnector
was usng 'Type' to log CipherDetail or BlobMetadata type information.
It causes 'fdbserver' binary to crash/restart

Patch addresses the usage to ensure TraceEvent compatibility is
maintained. Further, it proposes following test improvements:
1. Fix existing unit-test that 'should' have caught this issue.
2. Add more test coverage.
3. Add CODE_PROBE to ensure relevant corner cases are validated.

Testing

RESTKmsConnectorUnit

(cherry picked from commit 7e872c4)
…ecause otherwise we could delete our coordinated state (apple#10136)
sfc-gh-etschannen and others added 23 commits August 15, 2023 17:24
* fix: Non-storage processes were not being checked for locality exclusions
fix: Data distribution when not detect a newly added process was locality excluded
fix: RemoveServerSafely did not wait for processes to be excluded before killing them when excluding localities

* fix: do not allow locality based excludes if they cannot exclude the required addresses

* fix: do not allow rebootAndDelete when doing locality based excludes because it changes the processId of a process
* Prevent Status actor from bubbling up timeout error

Description

Patch addresses occurrences where Status.actor ends up bubbling timeout error
up to the ClusterController causing recovery to be triggered when
ClusterGetStatus timeout for some reason.

Testing

devRunCorrectness - 100K

* Update fdbserver/Status.actor.cpp

Co-authored-by: A.J. Beamon <aj.beamon@snowflake.com>

---------

Co-authored-by: A.J. Beamon <aj.beamon@snowflake.com>
…ectly deal with re-aligning keys after downsample (apple#10794)
…pple#10811)

* Normalize clear costs for quota throttler

* Track cost of point clears
Description

Release notes - snowflake 71.3.4

Testing
…r a request (apple#10808)

* FDBCORE-5950: Ensure exponential backoff while refreshing KMS URLs for a request

Description

KMS server(s) outage could last for longer times, RESTKmsConnector
follow exponential-backoff semantics when retrying request to
same KMS server, however, it doesn't follow exponential backoff
semantics when attempting to retry request after refreshing
KMS URLs by reading on-disk discovery file.

Patch addresses the concern by ensuring request follow exponential
backoff semantics for above mentioned scenario.

Testing

devRunCorrectness - 100K 20230824-141642-ahusain-6351198682ff45b7

* Update the initial and max delays as KNOBs

Description

Testing
…ake/release-71.3) (apple#10830)

* Add trace details about over-replicated shard and range count

* Fix count refresh
…) (apple#10832)

* EaR: Handle KMS timeout in storage server and commit proxy

* EaR: reduce KMS request timeout
Description

KMS health check monitor is instrumental in determing
FDB <-> KMS connectivity issues (no connection or flakiness).
Patch lower the check interval to be under FDB commit timeout
to get assist diagnosing such issues.

The change do translate to increased request to KMS, however,
the health-check asks for FDB system keyspace encryption key,
it would be hitting cache on KMS server all the time (unless
a KMS server got replaced and the first read is a miss)

Testing

devRunCorrectness - 100K
Co-authored-by: Ata E Husain Bohra <ata.husain@snowflake.com>
Description

Bump version to 71.3.5

Testing
Description

Release notes for snowflake/71.3.3

Testing
DatabaseContext currently leaks memory by creating `Counter`s with
unique IDs on construction. Each status json call creates a new
`DatabaseContext` object, causing a memory leak over time.
Description

Bump snowflake/release-71.3 from 71.3.5 -> 71.3.6

Testing
* Release notes for snowflake/release-71.3.5

Description

Release notes for snowflake/release-71.3.5a

Testing
Description

Release notes - snowflake/release-71.3.6

Testing
@mo74m3ed

Copy link
Copy Markdown
Owner Author

@copilot resolve the merge conflicts on this branch.

@mo74m3ed

Copy link
Copy Markdown
Owner Author

@copilot resolve the merge conflicts on this branch.

1 similar comment
@mo74m3ed

Copy link
Copy Markdown
Owner Author

@copilot resolve the merge conflicts on this branch.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.