-
Notifications
You must be signed in to change notification settings - Fork 3.8k
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.
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
roachtest: sqlsmith/setup=seed/setting=no-ddl failed #62846
Comments
The panic is coming from cockroach/pkg/sql/sem/builtins/builtins.go Lines 3828 to 3843 in a49cd0a
Looks like A side note to myself: I think we should be catching panics coming from |
62498: utilccl,kvccl: improve performance when checking enterprise features r=tbg a=erikgrinaker **utilccl: cache license decoding** Previously, the `utilccl` package would decode the license from the the base64-encoded Protobuf representation in settings every time it was needed, which was sufficient for its uses. However, recently there's been a need to check whether enterprise features are enabled in hot paths (e.g. with follower reads as seen in #62447), making the decoding cost too great. This patch adds `cluster.Settings.Cache` as a shared cache, and uses it to cache decoded licenses with a private key type. **utilccl,kvccl: add IsEnterpriseEnabled for faster license checks** `utilccl.CheckEnterpriseEnabled()` is used to check whether a valid enterprise license exists for a given feature. If no valid license is found, it returns an error with specific details. However, `kvccl` used this function in follower read hot paths, and instantiating an error when follower reads are unavailable could have significant overhead -- see e.g. #62447. This patch adds `IsEnterpriseEnabled()`, which has the same behavior as `CheckEnterpriseEnabled()` but returns a boolean instead. This is significantly faster since we can avoid instantiating a custom error each time. `kvccl` is also updated to use this in hot paths. Resolves #62489. Release note: None 62642: colserde: fix the edge case with nulls handling r=yuzefovich a=yuzefovich When serializing the data of Bool, Bytes, Int, and Float types when they don't have any nulls in the vector, we don't explicit specify the null bitmap. Previously, when deserializing such vectors with no nulls we would simply call `UnsetNulls` on the `coldata.Nulls` object that is currently present. However, it is possible that already present nulls object cannot support the desired batch length. This could lead to index out of bounds accesses. Note that in the vast majority of cases this likely doesn't happen in practice because we check `MaybeHasNulls`, and that would return `false` making us omit the null checking code. Fixes: #62636. Release note (bug fix): Previously, CockroachDB could encounter an internal error in rare circumstances when executing queries via the vectorized engine that operate on columns of BOOL, BYTES, INT, and FLOAT types that have a mix of NULL and non-NULL values. 62740: workload: add idle-conns flag for adding idle connections to tpcc r=rafiss a=RichardJCai workload: add idle-conns flag for adding idle connections to tpcc Release note: None #62526 62814: tenantrate: add "test" that reports IOPS estimations r=RaduBerinde a=RaduBerinde This change adds a "test" facility which takes the description of a uniform workload (read percentage, read size, write size) and prints out an estimation of the sustained IOPS and burst IO. This will allow a better understanding of how changes to the settings or the mechanism translate into IOPS changes. Release note: None 62833: kvserver: deflake TestFollowerReadsWithStaleDescriptor r=aayushshah15 a=aayushshah15 A preceding change (#62696) introduced a flakey update to this test. Prior to that change, this test was using 2 voting replicas but that change tried to make it use 1 voter and 1 non-voter instead (as a litmus test for the new syntax added in #62696). The test rebalances a replica away from a node and ensures that a historical read sent immediately afterwards gets re-routed to the leaseholder replica, since the receiving store had its replica destroyed. However, when we're using a non-voter in this test, that non-voter may not have learned about this replication change by the time it receives this historical query and that fails the assertion. This commit re-organizes the test and fixes the flake. Release note: None 62862: testutils: add skip.UnderBazelWithIssue r=rickystewart a=stevendanna This is to skip individual tests under bazel. This seems a bit more fine-grained than the broken_in_bazel tag in the bazel configuration but also allows us to avoid skipping tests that work outside of bazel in our main test suite. Release note: None 62877: Added CACHE to SEQUENCE syntax diagrams r=ericharmeling a=ericharmeling Follow-up of #56954. Release justification: non-production code changes Release note: None 62889: colexecerror: catch panics from packages in sql/sem folder r=yuzefovich a=yuzefovich Previously, we would only catch panics from `sql/sem/tree` package. Recently sqlsmith encountered a crash because of a panic in `sql/sem/builtins` package, and I believe it is reasonable to catch panics from that package as well as from `sql/sem/transform`, so we will now be catching based on `sql/sem` prefix. Addresses: #62846. Release note: None 62898: build: install essential build tools in teamcity build agents r=jlinder a=rickystewart In #62815, we migrated from an alternative way of installing golang, the `longsleep/golang-backports` deb repo, to the currently recommended install method found at https://golang.org/doc/install -- namely, we download a tarball and then just unzip it in the right spot. This works perfectly, *except* that the deb package had a dependency on build tools like `gcc` and `make`, and certain build configurations had come to depend on their global installation (namely, all those that don't use `builder.sh` to run a build). This resulted in a couple of failures being reported: * https://teamcity.cockroachdb.com/buildConfiguration/Cockroach_ExampleORMs/2834741 * https://teamcity.cockroachdb.com/buildConfiguration/Cockroach_UnitTests_Acceptance/2834732 We just install [`build-essential`](https://packages.ubuntu.com/xenial/build-essential) here, which is the easiest way to get all of that stuff. Release note: None Co-authored-by: Erik Grinaker <[email protected]> Co-authored-by: Yahor Yuzefovich <[email protected]> Co-authored-by: richardjcai <[email protected]> Co-authored-by: Radu Berinde <[email protected]> Co-authored-by: Aayush Shah <[email protected]> Co-authored-by: Steven Danna <[email protected]> Co-authored-by: Eric Harmeling <[email protected]> Co-authored-by: Ricky Stewart <[email protected]>
(roachtest).sqlsmith/setup=seed/setting=no-ddl failed on master@9891b8db5ea2e9339a04fd2731a16658103e969f:
More
Artifacts: /sqlsmith/setup=seed/setting=no-ddl See this test on roachdash |
roachtest.sqlsmith/setup=seed/setting=no-ddl failed with artifacts on master @ 71a023f9212cf1285a4475ed85d5ab6eefa9bdf7:
Reproduce
To reproduce, try: #!/usr/bin/env bash
#!/bin/bash
set -euxo pipefail
# NB: invoke this script with "caffeinate" on OSX and/or linux to
# prevent runs failing due to standby.
sha=$(git rev-parse HEAD)
if [ ! -f roachtest.$sha ]; then
./build/builder.sh mkrelease amd64-linux-gnu bin/{roach{prod,test},workload}
mv -f bin.docker_amd64/roachprod roachprod.$sha
mv -f bin.docker_amd64/workload workload.$sha
mv -f bin.docker_amd64/roachtest roachtest.$sha
fi
if [ ! -f cockroach.$sha ]; then
./build/builder.sh mkrelease amd64-linux-gnu
mv cockroach-linux-2.6.32-gnu-amd64 cockroach.$sha
fi
# NB: consider adding --debug if it is useful to let the clusters
# for failed tests survive.
./roachtest.$sha run "sqlsmith/setup=seed/setting=no-ddl" \
--port 8080 --count 10 --cpu-quota 500 \
--roachprod roachprod.${sha} --workload workload.${sha} \
--cockroach ./cockroach.$sha \
--artifacts artifacts.$sha | tee roachtest-stress.${sha} Same failure on other branches
Internal log
|
dup of #63819 |
@yuzefovich this latest failure looks unrelated to #63819, since the error is "error: index out of range [1] with length 1" and crdb_internal.create_join_token() is not part of the query. Is there already another issue open for that error? |
The latest one fixed by #63446:
|
(roachtest).sqlsmith/setup=seed/setting=no-ddl failed on master@d145e9fc02064a8b6b4179b5af7da5238b192f74:
More
Artifacts: /sqlsmith/setup=seed/setting=no-ddl
See this test on roachdash
powered by pkg/cmd/internal/issues
The text was updated successfully, but these errors were encountered: