workload: add indexes workload#35077
Conversation
|
This change is |
ajwerner
left a comment
There was a problem hiding this comment.
This is a fine start. I can imagine we'll extend this by adding control for the distribution of writes.
Reviewed 1 of 3 files at r1.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @ajwerner and @nvanbenschoten)
pkg/workload/indexes/indexes.go, line 15 at r1 (raw file):
// permissions and limitations under the License. package bank
package indexes?
pkg/workload/indexes/indexes.go, line 69 at r1 (raw file):
New: func() workload.Generator { g := &indexes{} g.flags.FlagSet = pflag.NewFlagSet(`indexes`, pflag.ContinueOnError)
should we add an init flag to be parallel to kv?
pkg/workload/indexes/indexes.go, line 73 at r1 (raw file):
g.flags.IntVar(&g.idxs, `secondary-indexes`, 1, `Number of indexes to add to the table.`) g.flags.BoolVar(&g.unique, `unique-indexes`, false, `Use UNIQUE secondary indexes.`) g.flags.IntVar(&g.payload, `payload`, 64, `Size of the payload only present in the primary key.`)
only present in the primary index? I found this documentation sort of confusing.
danhhz
left a comment
There was a problem hiding this comment.
Reviewable status:
pkg/workload/indexes/indexes.go, line 103 at r1 (raw file):
} for i := 0; i < w.idxs; i++ { ddl := fmt.Sprintf(`CREATE%s INDEX idx%d ON indexes (col%d)`, unique, i, i)
unless you're specifically interested in testing schema changes, it's usually better to put these in the table schema. then if you're dealing with fixtures, you get them during the IMPORT
Closes cockroachdb#34912. This adds a new workload called `indexes`. Unlike `kv`, the workload is specifically designed to stress secondary indexes. Its schema looks like: ``` CREATE TABLE indexes ( key UUID NOT NULL PRIMARY KEY, col0 INT NOT NULL, col1 INT NOT NULL, col2 INT NOT NULL, col3 INT NOT NULL, col4 INT NOT NULL, col5 INT NOT NULL, col6 INT NOT NULL, col7 INT NOT NULL, col8 INT NOT NULL, col9 INT NOT NULL, payload BYTES NOT NULL ) ``` and it includes knobs to specify how many of the "col" columns should be indexed and whether they should be indexed using a UNIQUE specifier. The workload also includes a knob to specify the size of the `payload` column. Release note: None
6a90e28 to
74e3029
Compare
nvb
left a comment
There was a problem hiding this comment.
Reviewable status:
pkg/workload/indexes/indexes.go, line 15 at r1 (raw file):
Previously, ajwerner wrote…
package indexes?
Done.
pkg/workload/indexes/indexes.go, line 69 at r1 (raw file):
Previously, ajwerner wrote…
should we add an
initflag to be parallel to kv?
The init flag is part of workload. See pkg/workload/cli/run.go.
pkg/workload/indexes/indexes.go, line 73 at r1 (raw file):
Previously, ajwerner wrote…
only present in the primary index? I found this documentation sort of confusing.
How's that?
pkg/workload/indexes/indexes.go, line 103 at r1 (raw file):
Previously, danhhz (Daniel Harrison) wrote…
unless you're specifically interested in testing schema changes, it's usually better to put these in the table schema. then if you're dealing with fixtures, you get them during the IMPORT
Done.
ajwerner
left a comment
There was a problem hiding this comment.
Reviewed 1 of 3 files at r1, 1 of 1 files at r2.
Reviewable status:
|
bors r+ |
35077: workload: add indexes workload r=nvanbenschoten a=nvanbenschoten Closes #34912. This adds a new workload called `indexes`. Unlike `kv`, the workload is specifically designed to stress secondary indexes. Its schema looks like: ``` CREATE TABLE indexes ( key UUID NOT NULL PRIMARY KEY, col0 INT NOT NULL, col1 INT NOT NULL, col2 INT NOT NULL, col3 INT NOT NULL, col4 INT NOT NULL, col5 INT NOT NULL, col6 INT NOT NULL, col7 INT NOT NULL, col8 INT NOT NULL, col9 INT NOT NULL, payload BYTES NOT NULL ) ``` and it includes knobs to specify how many of the "col" columns should be indexed and whether they should be indexed using a UNIQUE specifier. The workload also includes a knob to specify the size of the `payload` column. Release note: None 35116: sql: add support for extra session vars for pg compatibility r=knz a=knz Fixes #35109. This adds compatibility support for the following variables, with only the default values that make sense in CockroachDB: - `row_security` - `synchronize_seqscans`, - `lock_timeout`, - `idle_in_transaction_session_timeout` Release note: None 35134: storageccl: leave params in workload URIs r=dt a=dt workload URIs do not contain anything sensitive and including them in the job make it easier see what it was actually doing. Release note: none. 35151: sql: tolerate non-existent databases for plan cache invalidation r=knz a=knz Fixes #35145. Release note (bug fix): CockroachDB again properly reports when a database used during PREPARE does not exist any more when EXECUTE is used. 35162: storage: remove TODOs in cmd_push_txn r=nvanbenschoten a=nvanbenschoten This was discussed in #33523 (review) but I missed digging into it at the time. I still don't think I have my head fully wrapped around it, but I do enough to know that it's not something I want to start pulling on right now. Release note: None Co-authored-by: Nathan VanBenschoten <nvanbenschoten@gmail.com> Co-authored-by: Raphael 'kena' Poss <knz@cockroachlabs.com> Co-authored-by: David Taylor <tinystatemachine@gmail.com>
Build succeeded |
Closes #34912.
This adds a new workload called
indexes. Unlikekv, the workload isspecifically designed to stress secondary indexes. Its schema looks
like:
and it includes knobs to specify how many of the "col" columns should be
indexed and whether they should be indexed using a UNIQUE specifier. The
workload also includes a knob to specify the size of the
payloadcolumn.
Release note: None