fix #2555: add backreference in CQs #3876
Conversation
dgnorton
force-pushed
the
dgn-cq-backrefs
branch
from
ff6640a to
056e013
Compare
|
Looks like Circle blew up due to memory pressure. I've seen this before, and it can happen if the code allocates huge buffers by mistake, perhaps in a loop. |
| @@ -699,6 +702,8 @@ func (p *Parser) parsePrivilege() (Privilege, error) { | |||
| // parseSelectStatement parses a select string and returns a Statement AST object. | |||
| // This function assumes the SELECT token has already been consumed. | |||
| func (p *Parser) parseSelectStatement(tr targetRequirement) (*SelectStatement, error) { | |||
| fmt.Println("parseSelectStatement: start") | |||
|
Can you add a test to ensure that |
|
@dgnorton -- did you work out what caused Circle to fail due to memory pressure? |
| @@ -346,7 +351,11 @@ func (s *Service) runContinuousQueryAndWriteResult(cq *ContinuousQuery) error { | |||
| } | |||
|
|
|||
| if s.loggingEnabled { | |||
| s.Logger.Printf("wrote %d point(s) to %s.%s.%s", len(points), cq.intoDB(), cq.intoRP(), cq.Info.Name) | |||
| db := cq.intoDB() | |||
There was a problem hiding this comment.
This is purely a logging thing? This looks like logic that should be in the cq object, no?
|
@otoolep I added a |
dgnorton
force-pushed
the
dgn-cq-backrefs
branch
2 times, most recently
from
5b550d7 to
a3a38ad
Compare
Add new query syntax to allow the following in CQs: INTO "1hPolicy".:MEASUREMENT
dgnorton
force-pushed
the
dgn-cq-backrefs
branch
from
a3a38ad to
2739f5e
Compare
| exp: `{"results":[{"series":[{"name":"cpu","columns":["time","count","host","region","value"],"values":[["` + interval2.UTC().Format(time.RFC3339Nano) + `",3,null,null,null]]},{"name":"gpu","columns":["time","count","host","region","value"],"values":[["` + interval1.UTC().Format(time.RFC3339Nano) + `",2,null,null,null],["` + interval0.UTC().Format(time.RFC3339Nano) + `",1,null,null,null]]}]}]}`, | ||
| params: url.Values{"db": []string{"db0"}}, | ||
| }, | ||
| // TODO: restore this test once this is fixed: https://github.com/influxdb/influxdb/issues/3968 |
There was a problem hiding this comment.
Is there anything unique about this test other than duplicate tags? Also, the idiom I like to use is put a skip in my test struct, then the code isn't commented out, and if you do a verbose log, you can t.Logf out any skipped table tests.
There was a problem hiding this comment.
It checks the results of a CQ with group by *.
There was a problem hiding this comment.
Ok, that's what I figured, just wanted to check.
|
This might be beyond the scope of this PR, but how do we handle:
Looks good. +1 |
I mocked enough of the interfaces in the CQ service tests that it should be relatively easy to add tests for those things but I don't think they're currently tested. |
fix #2555: add backreference in CQs
|
I want to add suffix for measurement names ,may by like that |
Add new query syntax to allow the following in CQs:
In this example, the regular expression in the
FROMclause would expand to all measurements and the points selected from each of those measurements would be written into the"1hPolicy"retention policy with the same measurement names.