Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Improve tsdump functionality and performance #75993

Closed
rimadeodhar opened this issue Feb 3, 2022 · 0 comments
Closed

Improve tsdump functionality and performance #75993

rimadeodhar opened this issue Feb 3, 2022 · 0 comments
Assignees
@rimadeodhar
Labels
C-enhancement Solution expected to add code/behavior + preserve backward-compat (pg compat issues are exception)

Comments

@rimadeodhar
Copy link
Collaborator

rimadeodhar commented Feb 3, 2022 •
edited by cockroach-jira-scripts
Loading

tsdump performs strict checks that verify that it has data for exactly the stores it expects to see according to the node->store mapping. It can happen that there legitimately isn't data there, for example when nodes are down during the tsdump window. We still want the strict checks, but now tsdump will should defer returning an error and allow restarting with a - file to display the data anyway.

Cluster settings are should be set via SET CLUSTER SETTING as opposed to overriding settings directly.

Jira issue: CRDB-12908
@rimadeodhar rimadeodhar added C-enhancement Solution expected to add code/behavior + preserve backward-compat (pg compat issues are exception) T-observability-inf labels Feb 3, 2022
@rimadeodhar rimadeodhar self-assigned this Feb 3, 2022
@rimadeodhar rimadeodhar mentioned this issue Feb 3, 2022
Closed
@rimadeodhar rimadeodhar mentioned this issue Mar 1, 2022
Merged
craig bot pushed a commit that referenced this issue Mar 8, 2022
@AlexTalks @rickystewart
@rimadeodhar @rafiss
Merge #76611 #76897 #77247 #77483 #77493
f15acd1 
76611: roachpb: add lock table metadata structures r=AlexTalks a=AlexTalks

This change adds the protobuf structures, as well as the method, needed
to capture the state of a replica's lock table.  This is part of the
work coming out of #75541, and is needed to be able to implement the
`QueryLocks` RPC.

Release justification: (Category 2) The structures created here are
entirely new, for the purposes of observability, and do not represent a
risk to any existing code or functionality.

Release note: None


76897: release: use bazel to build release binaries r=rail,jlinder a=rickystewart

This is a pretty large swath of related changes that update the release
pipeline to use Bazel for builds.

1. Update `.bazelrc` to add a new `*base` config for each cross-config.
   The existing cross configs will continue to work as before. Each
   `*base` config has the same contents as the corresponding cross
   config EXCEPT the `--workspace_status_command` argument is missing.
2. Stamping. `build/bazelutil/stamp.sh` now takes up to 3 more
   arguments in addition to the target triple: `build-channel`,
   `build-tag`, and `build-type`. These are documented in the script and
   have appropriate defaults for development builds if not specified.
3. Refactor `pkg/release`; the new API is more high-level rather than
   allowing injecting arbitrary `gotags`, etc. In the future it won't
   be impossible to add this functionality back in if we want it.
4. Prepare to fork the `Make and Publish Build` build configuration in
   TC by moving its script into `build/teamcity` and update the script
   to use Bazel.
5. Update `bazci` to make sure it reads the Windows `geos` libraries
   from the appropriate location in `bazel-bin`.
6. Update `publish-artifacts` and `publish-provisional-artifacts` to use
   the new `pkg/release` API.

Closes #67147.
Closes #67161.
Closes #67279.
Closes #67322.

Release note (general change): Use Bazel to build release binaries
Release justification: update release process

77247: server: Allow bypass of tsdump's strict checks r=abarganier a=rimadeodhar

tsdump performs strict checks that verify that it
has data for exactly the stores it expects to see
according to the node->store mapping. It can happen
that there legitimately isn't data there, for example
when nodes are down during the tsdump window. We still
want the strict checks, but now tsdump will defer
returning an error and allow restarting with a "-"
in lieu of the tsimport file name  to display the data
anyway.

Additionally, cluster settings are set via SET CLUSTER SETTING
as opposed to overriding settings directly.

Release note (cli change): debug tsdump command allows
viewing timeseries data even in case of node failures
by rerunning the command with the import filename set
to "-".

Addresses #75993

Release justification: low risk, high benefit changes to existing functionality


77483: sql/catalog: add return error to RunPostDeserializationChanges r=ajwerner a=rafiss

Release justification: low risk enhancement to existing functionality.

Release note: None

77493: sql: make session_revival_token.enabled tenant-ro r=otan,jaylim-crl a=rafiss

I was hoping to wait for the new cluster setting syntax to be completed,
but since it's getting close to the branch cut time I'd rather merge
this now so we don't forget at the last minute.

Release justification: low risk change to new functionality.

Release note: None

Co-authored-by: Alex Sarkesian <[email protected]>
Co-authored-by: Ricky Stewart <[email protected]>
Co-authored-by: rimadeodhar <[email protected]>
Co-authored-by: Rafi Shamim <[email protected]>
@rimadeodhar rimadeodhar mentioned this issue Mar 16, 2022
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@rimadeodhar rimadeodhar

Labels
C-enhancement Solution expected to add code/behavior + preserve backward-compat (pg compat issues are exception)
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@rimadeodhar