Age | Commit message (Collapse) | Author | Files | Lines |
|
Small differences in Tx packet counts can cause
the previous way (which used transmit_rate,
computed from packets sent and nominal duration)
to report lower bounds as higher than upper bounds.
Also, reconf tests need the target (not actual) rate.
Change-Id: If5222dabd9762b8d9c2e22d875c12ad355bc6838
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Improve keyword Documentation.
+ Reorder measurements, higher loads first.
Change-Id: I1aeab1c8ada268422e9f421e075aff2ac9eec23f
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Without this, if the search part returns min_rate,
then reconf part reports few seconds with a small loss count.
But min_rate means there is a bug or misconfiguration,
so the tests should fail, instead of returning misleading results.
Change-Id: I566608b40ae11d6aaa3fe022a32da97901e1999c
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Latency measurements break for loads less than 9 kpps per direction.
Setting min_rate to 90 kpps prevents 10% latency trial from breaking.
90 kpps should be enough for any type of test.
When some test finds a lower value, this patch will make the fail
with a message saying the loss rate at minimal load is too high
(instead of a cryptic error from 10% measurement without this patch).
Change-Id: I97b8428aa5044da23343b86447857b3a101f6baa
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Before [0], the NDRPDR test ended with two measurements
that perform runtime stats collection, each 2 seconds long.
[0] then decreased that duration to one second,
and removed the PDR measurement.
This change re-adds the PDR measurement,
while keeping the duration at 1 second.
[0] https://gerrit.fd.io/r/c/csit/+/24427
Change-Id: I76828f11863a9e278cdd0cf60f2792d5b614ce43
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I46bd4b58bdd02a3deed38c158214e1e9977dde05
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: I250ffd196045bc623d5d38f30f16a08bcaea137c
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
|
|
+ Add both NDR and PDR telemetry capture
+ Speedup sockets
+ Adjust privileges
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: Ia6fd5d405e6fb410651d8b705c921653753aea10
|
|
Change-Id: I1392c06b1d64f62b141d24c0d42a8e36913b15e2
Signed-off-by: Jan Gelety <jgelety@cisco.com>
|
|
Change-Id: I8e0a14f59e20d61096e069bb80a0e2c00977ac5c
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
+ This will help unify driver base differences between TCs.
+ Decrease amount of logic needed for suite generator.
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I1e84ba361dc1e829f0612c58a61096e2633ce0c5
|
|
+ Enable hdrh in trex server.
+ Append hdrh coded output after min/avg/max/.
+ Read (not show nor decode) hdrh value in PAL.
+ Also, remove old ndrpdrdisc code.
Change-Id: I99d99f10386a621772b5419ca1f36080fa15aca7
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
TRex does not zero the server counters.
It copies the values to use as reference,
and subtracts them when asked for results.
But the reference is stored in the client (not the server).
And CSIT uses different scripts to start and stop async traffic,
which means different clients.
This patch introduces a workaround.
Async start will return xstats objects to use as reference,
and async stop will use the objects to compute the correct results.
The xstats objects are stored in TrafficGenerator instance.
Sync measurement does not export the counters, to shorten logs.
Other improvements:
+ Make stop_traffic_on_tg return measurement results directly.
+ Rename --async to --async_start as "async" is reserved in Python 3.7
+ Minor pylint, docstring and typo fixes.
Change-Id: I5fc56a0763afb7d62cfa7c0651f96b6867de3e15
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Rename bool unidirection to int traffic_directions.
+ Rename "untagged" to "initial" for bandwidth calculation.
+ Fix latency measurement for unidirectional traffic.
+ Remove duplicate colon in soak test message.
+ Edit PAL to accept both forms.
+ Fix minor documentation issues.
Change-Id: I6c76f2dc090ae493f2fbd7e9ccd45229d2306dea
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Ability to get stats from CNF via SocketPAPI
- Remove obsolete functions
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I4d1b32a7279244592be96644e4f8a530c4f29a15
|
|
Instead of using EnsureGlobalVariable,
which is clunky to use from Python.
As a consequence, all caps variables from Constants.py are used directly
and tests/__init__.robot and robot_enhancements.robot are deleted.
+ Rename the CRC global kill switch based on ci-man review.
Change-Id: I10723792475bc83352bf8c3b7f7946ecf885a194
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Ticket: CSIT-1551
+ Config copied from other suites.
+ Layer keywords prepared for additional chains.
+ Renamed "create" argument to "vlan_per_chain".
+ TrafficGenerator result extraction improvements.
+ Also unified TG type and subtype checking.
+ Throughput estimated as PDR for zero PLR.
+ Actual Qemu start is skipped.
+ Parse xstats instead of stats in trex stop scipt.
- Because stats are unconditionally cleared on each script connect.
+ Remember start values to get better ReceiveMeasurementResult at stop.
+ Expose loss count and time as appropriate test message.
+ Autogen support.
+ Multiple chain amounts and core densities.
- Only 1 additional chain.
+ Proper TAGs. Documented.
+ Clarified min framesize tags are for TG-DUT.
+ Added a TODO concerning overhead values.
+ Added 118B tag.
+ Robot performance keywords improvements.
+ Added missing return value documentation.
+ Frame size as required test variable, not an explicit argument.
+ MLRsearch keyword to store result (not into test message).
- TODO: De-duplicate: main keyword to measure, minor to set test message.
- But implementation uses the fact default arguments are different.
+ Keywords for starting and stopping background traffic.
- No checks whether start-stop pair is complete.
+ Multiple minor indentation and logging improvements.
- Missing / TODOs:
- Support also vlan_per_chain=True?
- Support dot1q without vxlan?
- Support also deletion?
- How to package config steps so -reconf can be generated from -ndrpdr?
Change-Id: If0270697f9ffa837c34ef802917220a53f156703
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Logic improvements for rls1908 and post-pylint fixes:
+ Reduce search time to 30 minutes.
+ Use average instead alternating loads.
+ Rework log/exp avg/stdev from two estimates.
+ Introduce and use pessimistic variance of dual trackers.
+ Introduce safe_exp to use when None does not skip code.
+ Use dot relative imports (instead of disabling pylint).
+ Complete docstrings for simpler functions.
+ Append docstrings to named tuples.
+ Somewhat unify docstrings related to the same arguments.
+ Slightly improve intentation.
+ State named tuples as rtype where used.
+ Add returns and rtype to __repr__ where missing.
+ Return what docstring say (or update them).
+ Explicit copying for vector and matrix values.
Change-Id: I884c68b4839c5df5e8bef95e463666599603a0ff
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
- Remove translate TC as they have no perf opposite
- Remove unused keywords
- Merge similar keywords / rename
- Move keywords from perf library to categories
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I6433006136e2497ad8fb17193aa20c578d836999
|
|
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I3721900b91a9c8e9db8766bdcb9771b6bcf95bca
|
|
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: Ic15a15c426db04defc15840883a772a0eb8e4885
|
|
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: Ib3620689acb3641b2ec9a49e2ce552bab3f00eef
|
|
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: Ideddcea56120cf770aec78ca24813baf7d05ba11
|
|
+ Move all base ipv6 related KW to library
+ Remove useless KW to save space in output.xml
+ Remove cross dependency in ip6.robot
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I8c5f72e9852c1017356ab6034c4d271cf7b53265
|
|
+ Move all base ipv4 related KW to library
+ Remove useless KW to save space in output.xml
+ Remove cross dependency in ip4.robot
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: If1e3a6370d36caf4ce7411fb956a0b7d0f939b6b
|
|
Change-Id: I0d3c925ea4a6896a0df98db6ddaf4238e6291bf1
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: I8c8ae1d7eed8cf28fc63ce42b2dc36dd16a5c0da
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
+ NF_density compatible
+ Baseline VM tests compatible
+ Scalability NxM
- 3n variant is missing in VXLAN and DOT1Q due to requests to have it
asymmetrical - TODO
Change-Id: Ia817d7be00622b594b451bc07756240d1a379e04
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
- From unknown reason VAT was more tolerant
Change-Id: Ia740843c67604a18de40d14cc448138b9f079b41
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
- Code not used
Change-Id: I312635f8038c6d4d4db21d93e43df6b3636feb34
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
* replace the current VM image with kernel img
* rework keyword usage to make it consistent with performance tests
* remove resources/libraries/robot/shared/qemu.robot as it's not used
anywhere anymore
Change-Id: Ia5bc19e9e6ed9af031e4d9b5c0c89431fb49fd33
Signed-off-by: juraj.linkes <juraj.linkes@pantheon.tech>
|
|
- Alignment with PERF
Change-Id: Id7b2cb91c88e44cf9265c3376330f6aa579abf7a
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Do not support returning unverified replies anymore.
Basically, ".get_replies().verify_replies()" is now just ".get_replies()".
This allows fairly large simplifications both at call sites
and in PapiExecutor.py
+ Rename get_dumps to get_details.
+ Introduce get_reply and get_sw_if_index.
+ Rename variables holding get_*() value,
+ e.g. get_stats() value is stored to variable named "stats".
+ Rename "item" of subsequent loop to hint the type instead.
+ Rename "details" function argument to "verbose".
+ Process reply details in place, instead of building new list.
- Except hybrid blocks which can return both list or single item.
- Except human readable text building blocks.
+ Rename most similar names to sw_if_index.
- Except "vpp_sw_index" and some function names.
+ Use single run_cli_cmd from PapiExecutor.
+ Do not chain methods over multiple lines.
+ Small space gain is not worth readability loss.
+ Include minor code and docstrings improvement.
+ Add some TODOs.
Change-Id: Ib2110a3d2101a74d5837baab3a58dc46aafc6ce3
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: Ibd3e99390bf004173d3d126946b95ffe9ba8f196
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: I9287d5ed6d7876b98add3fb3d523e9764094843b
Signed-off-by: Jan Gelety <jgelety@cisco.com>
|
|
- Use explicit parameters
Change-Id: I21cbc8cf13ec5e455b20c685c4427a53a655f71f
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: Ic06a0a65429680e6ecdc3f5288d091c2c2630921
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
+ do not log all ip_route_add_del api calls to PAPI history
in case of scale tests
+ fix for SRv6 proxy tests
+ support of more complex structures in api calls/replies
in PapiExecutor
+ update of vpp stable build to version 19.08-rc0~487-gb98dbb1
Change-Id: I9972e126b68ce1cb57f781e3d90ea9b3914de69e
Signed-off-by: Jan Gelety <jgelety@cisco.com>
|
|
+ Phase II - aligning FUNC/PERF setup.
+ 2n/3n compatibility of suite setup.
+ Same structure of testcase for DEV/PERF.
+ DEVICE tests are now running VPP startup config in same
way as PERF.
+ DEVICE is using COREKEEPER!
- DEVICE tests are running topology compute twice for backward
compatibility. TODO to remove.
- Still phase III needed to align TAGS, LIBRARIES and remove OLD
setup phase.
Change-Id: Ib6e1b0ebf4abebf3cbe73b4fda5cac953c2a7270
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
+ Phase I - aligning FUNC/PERF
Change-Id: I46b1e43687ddb29f5ed3c6335fe1baf21ec02822
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: I675f9664391c9f18f312773812f765f345983cc1
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
Change-Id: Id84cf9c3afd53260b5bbf658a6a3870aa0c91333
Signed-off-by: Jan Gelety <jgelety@cisco.com>
|
|
Change-Id: If4eb592e2718c4564f1f7929ca6ff91c02f074a6
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
- 5K
- 10K
Change-Id: Iaa2c77d07c3686e09fe327c663d7d7fc42024034
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: Ic493253dc29dad4fc7e6541e4b839ca5898a6a40
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: I10eeb4ee30a57712824e68176d92d1ecb5f0d1b0
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: I0062710d58996be767a852dc00545fedd60a5c72
Signed-off-by: Jan Gelety <jgelety@cisco.com>
|
|
+ Extract several tracker classes and use them in Integrator.
+ Apply next_rate workarounds to focus more on critical region.
+ Rewrite stretch function, as the previous implementation was wrong.
+ Rework logging:
++ Use injected trace() function in Integrator.
++ Inject function that skips trace logging on default initialization.
++ Use the same multiprocessing-safe passing, but separate queue.
+ Set duration to 120m in Robot, but keep at 30m in Python.
+ Apply minor changes to make computations more reproducible:
++ Attempt to log everything needed by reproducibility in one line.
++ Log samples used, and make it settable as upper limit.
++ Use repr output in TRex scripts, to avoid rounding when copypasting.
+- Numpy seems to be sharing PRNG between processes.
+-- An edit to disable one thread is needed for full reproducibility.
+-- Such an edit is prepared, but will not be merged anytime soon.
Change-Id: Icf56429b34ab6de2d18a12e6c676e2d242b3ba4c
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
Change-Id: If4b013ef4bf4bc8f857913f3847ab3b2e2aa5a5d
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|
|
Change-Id: Ibe52125089f39e0ff17ec607a3ed00c61d52ab8c
Signed-off-by: Peter Mikus <pmikus@cisco.com>
|