-
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: ycsb/A/nodes=3/cpu=32 failed #37378
Comments
SHA: https://github.com/cockroachdb/cockroach/commits/d554884a4e474cc06213230d5ba7d757a88e9e46 Parameters: To repro, try:
Failed test: https://teamcity.cockroachdb.com/viewLog.html?buildId=1279548&tab=buildLog
|
SHA: https://github.com/cockroachdb/cockroach/commits/8abb47a1c9795c1463183bc44e776b054bece682 Parameters: To repro, try:
Failed test: https://teamcity.cockroachdb.com/viewLog.html?buildId=1279687&tab=buildLog
|
SHA: https://github.com/cockroachdb/cockroach/commits/8abb47a1c9795c1463183bc44e776b054bece682 Parameters: To repro, try:
Failed test: https://teamcity.cockroachdb.com/viewLog.html?buildId=1279683&tab=buildLog
|
workload's Table schemas are SQL schemas, but cockroachdb#35349 switched the initial data to be returned as a coldata.Batch, which has a more limited set of types. (Or, in the case of simple workloads that return a []interface{}, it's roundtripped through coldata.Batch by the `Tuples` helper.) Notably, this means a SQL STRING column is represented the same as a BYTES column (ditto UUID, etc). This caused a regression in splits, which received some []byte data for a column tried to hand it to SPLIT as a SQL BYTES datum. This didn't work for the UUID column in tpcc's history table nor the VARCHAR in ycsb's usertable. Happily, a STRING works for both of these. It also seems to work for BYTES columns, so it seems like the ambiguity is fine in this case. When/if someone wants to add a workload that splits a BYTES primary key column containing non-utf8 data, we'll may need to revisit. A more principled fix would be to get the fidelity back by parsing the SQL schema, which in fact we do in `importccl.makeDatumFromColOffset`. However, at the moment, this hack works and avoids the complexity and the undesirable pkg/sql/parser dep. Closes cockroachdb#37383 Closes cockroachdb#37382 Closes cockroachdb#37381 Closes cockroachdb#37380 Closes cockroachdb#37379 Closes cockroachdb#37378 Closes cockroachdb#37377 Closes cockroachdb#37393 Release note: None
37401: workload: fix --splits regression introduced in #35349 r=tbg a=danhhz workload's Table schemas are SQL schemas, but #35349 switched the initial data to be returned as a coldata.Batch, which has a more limited set of types. (Or, in the case of simple workloads that return a []interface{}, it's roundtripped through coldata.Batch by the `Tuples` helper.) Notably, this means a SQL STRING column is represented the same as a BYTES column (ditto UUID, etc). This caused a regression in splits, which received some []byte data for a column tried to hand it to SPLIT as a SQL BYTES datum. This didn't work for the UUID column in tpcc's history table nor the VARCHAR in ycsb's usertable. Happily, a STRING works for both of these. It also seems to work for BYTES columns, so it seems like the ambiguity is fine in this case. When/if someone wants to add a workload that splits a BYTES primary key column containing non-utf8 data, we'll may need to revisit. A more principled fix would be to get the fidelity back by parsing the SQL schema, which in fact we do in `importccl.makeDatumFromColOffset`. However, at the moment, this hack works and avoids the complexity and the undesirable pkg/sql/parser dep. Closes #37383 Closes #37382 Closes #37381 Closes #37380 Closes #37379 Closes #37378 Closes #37377 Closes #37393 Release note: None Co-authored-by: Daniel Harrison <[email protected]>
SHA: https://github.com/cockroachdb/cockroach/commits/d554884a4e474cc06213230d5ba7d757a88e9e46
Parameters:
To repro, try:
Failed test: https://teamcity.cockroachdb.com/viewLog.html?buildId=1279552&tab=buildLog
The text was updated successfully, but these errors were encountered: