mirror of
https://github.com/apple/foundationdb.git
synced 2025-05-14 01:42:37 +08:00
17570b5b10
Testspec is currently very permissive in very misleading ways. In particular, the tester parser itself will swallow K=V settings and apply them at the test level, which breaks how a person would expect the scoping to work. Other settings apply to the entire simulation run globally, but appear to be workload specific. Even further, others affect simulation cluster creation or test harness behavior, but can again be set anywhere in a testspec. This changes testspec parsing to error if a setting that applies globally is anywhere but the top of the file, or if a setting that applies test-wide is applied to a workload instead of a test.
24 lines
562 B
Plaintext
24 lines
562 B
Plaintext
testTitle=AsyncFileReadTest
|
|
useDB=false
|
|
runSetup=true
|
|
clearAfterTest=false
|
|
|
|
testName=AsyncFileRead
|
|
testDuration=36.0
|
|
;testDuration=36000.0 ;for a very long test
|
|
numParallelReads=32
|
|
readSize=4096
|
|
unbufferedIO=true
|
|
sequential=false
|
|
fileName=testfile
|
|
fileSize=1000000000 ;1GB
|
|
;fileSize=7000000000 ;7GB
|
|
;fileSize=100377287000 ;80% on 128GB drives
|
|
;fileSize=141774800000 ;80% on intel drive
|
|
unbatched=true
|
|
writeFraction=0.33
|
|
uncachedIO=true
|
|
fillRandom=true
|
|
timeout=1000000000.0
|
|
;fixedRate=15000
|