Age | Commit message (Collapse) | Author | Files | Lines |
|
Each 5 VUs for searches uses a different filter:
no filter, module, property, cps path, and trust level.
This is needed since each search may have different
performance. This also helps reduce load versus doing
a combined module and property search all the time.
Issue-ID: CPS-2349
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I33a75260f97697d2061d10b80ba7eeb2f6c346f5
|
|
- run the test after the parallel cases
- add 30 seconds safe-period of time waiting for completion of
active threads (remained from previous test cases)
Issue-ID: CPS-2414
Change-Id: I1260b37cd4f1974eebcccf867b36172b407a9646
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
- as of now the cps limits given manually according to
the latest result
- there will be another commit will change the plot
job in cicd repo to reflect the values on the plots
Issue-ID: CPS-2350
Change-Id: Ic371f46894f1825dca1d26ec65ae0f62aa06c327
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
- stress the cps-and-ncmp module for batch read data operation
- for this aim use shared-ireations during async http request
- as parallel to this, consume expected number of messages
as fast as possible
Issue-ID: CPS-2268
Change-Id: I1b6724479dac7391cbb6407fda52d15176aa8985
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
De-registration is reported as 200,000 handles per second if all operations fail
- Added a running total for successful de-registrations of CM Handles
Issue-ID: CPS-2346
Change-Id: Ib4302d3b605ff1eaf4ecadac11b4e851a69210df
Signed-off-by: egernug <gerard.nugent@est.tech>
|
|
- Added unique IDs as metric tags for all the endpoints.
- Re-arranged order of public and prive js methods.
Issue-ID: CPS-2331
Change-Id: Ib876a647fb35110c50670c7222986e8a8a6f5ca0
Signed-off-by: sourabh_sourabh <sourabh.sourabh@est.tech>
|
|
Plot title say Module filter instead of Module and Property.
Issue-ID: CPS-2349
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ie23080a6085960b89132ad38d07d0c2b6f12471b
|
|
- Test load increased to 5 CM handle searches and 5 Id searches to
reflect current requirements.
- CM handles searches now use a combined search filter including
1 module and 1 public property, so we are testing with a heavier
load than required, for early warning of trouble.
- Test load reduced for passthrough operations to 4 VUs per operation
to reflect addition of batch operation.
- Added test of passthrough write operation with alternate ID.
- Additional refactoring.
Issue-ID: CPS-2349
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I88d2fe431f74821e8e0e976441efcf82aa320849
|
|
|
|
Instead of separate plots of test failures for each test case,
we have one single plot showing failures.
Issue-ID: CPS-2347
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ie79b75c34903a7df2037474b3536d4867c1a61e0
|
|
A change to DMI stub has broken k6 module search due
to module rename and increased registration time.
- increase setupTimeout by 2 minutes to factor in DMI startup delay
- module search uses module common to all module set tags
- only record times for searches returning number of CM-handles
Issue-ID: CPS-2353
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I9067dc73015cfb475cdce1e7a3cc2d5401449caf
|
|
- Changed k6 config to ignore all the error logs.
Issue-ID: CPS-2344
Change-Id: Iad7be43ae5bb909518f9f7d8e386a97c96c349e3
Signed-off-by: sourabh_sourabh <sourabh.sourabh@est.tech>
|
|
If an error occurs during k6 test, e.g. tcp connection refused, the
timing of failed response is included in the average response time.
This can lead to incorrect plots when errors occur, e.g. operations
appear to be faster when really the operations are failing.
- use custom Trend metrics for all measurements
- only log results when operations return expected response codes
Issue-ID: CPS-2345
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Iab1db5e5ca9530f33da239187348cabcd0605648
|
|
|
|
|
|
- install docker-compose in csit and k6 tests
- install k6 with kafka extension in k6 tests
Issue-ID: CPS-2376
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I8a67813a96dc7f321d4abe0aa91e236fd1060f3e
|
|
- add request failures to plot for each tests
- simplify setup and teardown function
Issue-ID: CPS-2347
Change-Id: I81226518af7bda12b0489ec46de08e27c96cc16d
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
Presently, k6 tests are using old cached DMI stub version in the
Jenkins job. This commit ensures latest images are pulled by
cleaning old images if a variable is set.
Issue-ID: CPS-2368
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I66a749c7cdeb7daee06a00b06dff50ead6854f28
|
|
The teardown script in k6 shuts down ALL docker containers when done.
This can cause failures in the groovy tests when running in parallel,
as the groovy test container for DB also shuts down.
This fix is to shut down using docker-compose.
Issue-ID: CPS-2367
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ic55ca6396dfc3170e4a326f34214d3a819977f50
|
|
Reported rates for passthrough operations is incorrectly reported,
due to it being averaged over whole test time including setup and
teardown. It was agreed to simply not report rates, and only report
NCMP overhead.
- remove http request rates for passthrough operations
- change VUs from passthrough read so that there are 10 total
Issue-ID: CPS-2264
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ida498955e69dffb332d6efeb8a8d73343565f9a1
|
|
Issue-ID: CPS-2357
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ibbd32282608ff707c8030d43df3cf85043242218
|
|
Issue-ID: CPS-2357
Change-Id: I21b980b1e36988d3b5daf07e4f6a14fe4309cd20
Signed-off-by: emaclee <lee.anjella.macabuhay@est.tech>
|
|
Issue-Id: CPS-2279
Change-Id: Iade53a85afb183c0df259c5e7fc12fa0f6986218
Signed-off-by: seanbeirne <sean.beirne@est.tech>
|
|
|
|
- add kafka support to k6 codebase
- add two new tests: data operation batch passthrough read
and data operation async batch read (from kafka)
- calculate the events/second via Trend
- add kafka ui support to docker-compose
Note: Before this commit being merged, the k6 executable
should have been compiled with the kafka extension of xk6-kafka
Issue-ID: CPS-2268
Change-Id: Ib7777b7bc9f15b210ea36d3541cba0e0c943f883
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
The POST operation returns 201, not 200.
This causes checks to fail when running k6 manually.
Issue-ID: CPS-2267
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I3c0e771b89aa09e41c9dbdf1c532599e1ab5440d
|
|
Measure NCMP overhead by subtracting DMI delay from request duration.
Issue-ID: CPS-2269
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ieb4758e9f7cab119da4c28e4a3febfa1361b2613
|
|
- change threshold from avg duration to req per second
- the delay for dmi has been hanlded in seprate commit
in dmi repository
- add delays through docker-compose env variables
Issue-ID:CPS-2285
Change-Id: I35978b21c208fdb4209b1610160dc5545ee22965
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
- the delay for dmi has been hanlded in seprate commit
- for simplicity only the POST request is considered
Issue-ID: CPS-2267
Change-Id: I62ede04dd8e2744fc723566742396df31c42ea5d
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
|
|
As per characteristics requirements document:
- measure registration in CM-handles/second
- measure deregistration in CM-handles/second
- summary table includes test case number, description,
units of measurement, actual value and limit.
Issue-ID: CPS-2269
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I838004da1c230ab722f49c2adacf34e730d7ac79
|
|
- this change for two test suits: CSIT and K6
- increase cps-and-ncmp replicas to 2
- introduce nginx as load-balancer
- set proxies inside nginx for the cps-and-ncmp URI paths
- upgrade hazelcast-spring to 5.3.7
Issue-ID: CPS-2038
Change-Id: If1b6c1e1218af16b24cc061507ea9b25d277d143
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
- Remove the tests that are not in the Characteristics
- Set .625 seconds/operation for cm handle id search with module filter
- Set 13 seconds/operation for cm handle search with module filter
Issue-ID: CPS-2286
Change-Id: I88d5773abadb29ce46914029332384ae8e55e032
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
To account for JVM warmup effect, the k6 test suite is run twice:
initially, the whole test suite is run without recording results.
- Added environment variables to allow setting total CM-handles and
batch size for registration. By reducing total CM-handles during
warm-up, many more operations will be run.
Issue-ID: CPS-2264
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ib1d1e76d31719d788a3e026797faa0a425b320a2
|
|
This commit moves all common request logic into a common folder.
It is needed to avoid duplication before adding JVM warmup phase.
- move registration-related code into common folder
- move passthrough operations into common folder
Issue-ID: CPS-2208
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ia9ebf61d21044b43063bde153f9c526e67d607c8
|
|
A cleanup of the k6 tests to avoid code duplication
Issue-ID: CPS-2208
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: If1e58ee87300cd28ed72b365d9a86e2709dc6c13
|
|
- Adjust timings based on previous runs.
- Reduce from 5 to 4 parallel clients for CM handle searches
to avoid OutOfMemoryErrors.
Issue-ID: CPS-2208
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: I339cc3df84b2295bb11371ab73b4948f43615c15
|
|
- add a generic template to print summary of k6 test results
- remove container logs since its around 50 mb
Issue-ID: CPS-2215
Change-Id: I8971fc30b9dc8be2ed16eda6755d17a91f608b48
Signed-off-by: halil.cakal <halil.cakal@est.tech>
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Signed-off-by: halil.cakal <halil.cakal@est.tech>
|
|
This commit adds K6 tests for NCMP functionalities:
- Registration of 20000 CM-handles
- Synchronous single CM-handle passthrough read operation
- CM-handle and ID searches for no filter, module, and public property
- De-registration of 20000 CM-handles
- A mixed load test
Also includes a README and shell scripts to run all tests.
Issue-ID: CPS-2199
Signed-off-by: danielhanrahan <daniel.hanrahan@est.tech>
Change-Id: Ia2e59674057ed28f5f9ceed9465133f7c0ce8318
|