distributed-database

What happened?

In a Jepsen test run of five etcd 3.5.3 nodes, with process pauses (sending processes SIGSTOP and SIGCONT), compaction (performed via the admin API), and defragmentation (via etcdctl), one etcd process crashed with the following error message:

{"level":"panic","ts":"2022-06-13T14:11:27.870Z","caller":"etcdserver/apply.go:638","msg":"unexpected error during txn","error

Enhancement

We have many place to use the exp/slices to sort our struct. It can improve performance by using pdqsort and clean branch check. But We have many placement using the old sort. Please remove it. Here is the example .

BTW, You haven't to change the bazel.build. it can be updated by robot.

• #36061
• #360

We have metrics recoding flush related stats (number of flushes, duration), but
those only work when issuing explicit Flush request.

Some sinks (cloud storage) also flush based on other criteria (size) -- we should
add metrics to record those flushes as well.

Jira issue: CRDB-17659

After DataSourcePreparer implementations use PipelineDDLGenerator, some classes are not used any more, they could be removed.

List:

• ActualTableDefinition
• TableDefinitionSQLType

Our CMake build process produces tarball packages using the cpack TGZ target. This produces a server and client package with a file layout meant to match where the various artifacts go.

Meanwhile, the packages we include on the download page are a flat archive of various binaries produced by the build process. This include one regular package and one debug package with debug symbols for a few

Jira Link: DB-3034

Description

Rocksdb Rate limiter supports specifying the granularity at which the rate is enforced. This defaults to 100ms

extern RateLimiter* NewGenericRateLimiter(
    int64_t rate_bytes_per_sec,
    int64_t refill_period_us = 100 * 1000,
    int32_t fairness = 10);

However, Docdb doesn't seem to be spec

Use case:

PosgreSQL client compatibility (e.g. Trino)
SQL compliance

CrateDB

CREATE TABLE doc.x (txt TEXT)
SELECT crate.doc.x.txt FROM crate.doc.x

PostgreSQL

postgres=# CREATE TABLE public.x (txt TEXT);
CREATE TABLE
postgres=# SELECT postgres.public.x.txt FROM postgres.public.x;
 txt 
-----
(0 rows)

[From PostgreSQL docs](https://www.postg

Enhancement

In SR MPP schedule mechanism, it is time-consuming to delivery the fragment instances of a query to BEs if the query has lots of Fragments and Fragment Instances. so we can add a instance delivery time metric for queries, this metric has the same lifetime as the QueryContext, so it can be put inside the QueryContext. the value of the metric will be the time duration from the fir

Now we have enabled our regress test in CI. We are planning to enable more test cases.

PG has a lot of existing test cases.
However:

1. We are not planning to support all of the features invoked in them.
2. It takes too much effort to be compatible with PG in terms of error messages.
3. We want to prioritize those SQLs that are more likely to be used by our users.

• advisory_lock.

Description

Some common traits don't bound by Send + Sync in definition but are required (every place) where they are used, like WalManager.

Proposal

<

Is there an existing issue for the same feature request?

• I have checked the existing issues.

Is your feature request related to a problem?

Several showcase SQL requires the support of this function:


select DATE_FORMAT(b.measurement_time,'%Y-%m-%d') 

select DATE_FORMAT(FROM_UNIXTIME(tm.create_time), %Y%m%d)

Describe the feature you'd like