74e6e135e9
I was trigger-happy with changes to qps_worker in the hope of improving
...
stability. I believe that this change was unneeded and actually
may hurt matters by holding a lock for too long.
Partially undoes #5444 , particular commit e7042b5
2016-06-08 13:32:16 -07:00
aaee544d69
merge and resolve conflicts
2016-06-03 15:34:01 -07:00
17487f9a22
Make grpc++ independent of protobuf
2016-06-03 15:21:15 -07:00
d4aa7cf5a5
Fix negation
2016-05-31 13:59:25 -07:00
1adb13143f
Add comment
2016-05-31 13:04:32 -07:00
dbd03b5e2d
Make client_channels be across all clients, not per-client
2016-05-31 13:00:33 -07:00
711472f846
Remove unused gflags and gtest includes
2016-05-31 12:57:38 -07:00
d8364f7af5
Deprecate perf_db use
2016-05-31 12:54:04 -07:00
e0ddc35aa3
Speed up tests
2016-05-23 15:57:48 -07:00
825cd4533c
Cleanup redundant tests
2016-05-23 15:54:38 -07:00
e67f7b6b54
Fix timeout on async server shutdown
...
Server was continuing to make requests for new calls forever, which were
starving out the shutdown sequence. Change order and win.
2016-05-23 14:50:24 -07:00
c0cb23b885
Merge branch 'guess_cpu' into test_affine
2016-05-20 11:17:31 -07:00
bd24a4611f
Better guesses at benchmarks turned unit tests cpu cost
2016-05-20 11:15:58 -07:00
c3b88b079e
Better testing
2016-05-16 11:21:17 -07:00
52237e3a96
clang-format
2016-05-13 14:37:25 -07:00
63d24f1c7d
Make qps workers connect before even warmup stage begins
2016-05-13 14:36:09 -07:00
b85e9d43f0
fix formatting
2016-05-11 10:21:45 -07:00
a21c7e9207
add support for scenario categories
2016-05-05 18:10:35 -07:00
e88030a2c2
Merge pull request #6340 from ctiller/fail_fast_breaks_qps
...
Disable fail fast for qps driver
2016-05-02 14:26:52 -07:00
97244d5810
clang-format, make code more self-documenting
2016-04-29 08:05:57 -07:00
5c123fd22b
Add a shard channel arg to client channel construction
...
C core automatically shares subchannels between channels. Therefore our
multiple channel performance tests were really testing single channel
performance.
2016-04-28 20:48:24 -07:00
b6e4440880
Disable fail fast for qps driver
2016-04-28 08:21:52 -07:00
241dea56c8
Merge pull request #6147 from ctiller/strong-includes
...
Rollup of changes from latest import
2016-04-21 18:59:06 -07:00
7beea147e5
Merge github.com:grpc/grpc into strong-includes
2016-04-21 17:07:36 -07:00
dfdfe26958
Make sure that there is at least one scenario, otherwise indicates malformed
...
input
2016-04-21 11:38:31 -07:00
8909428a1a
clang-format
2016-04-20 14:21:30 -07:00
92abf1cfd4
Merge branch 'master' into qps_driver_youre_fired
2016-04-20 13:06:25 -07:00
1f13c820ed
Eliminate unused source file
2016-04-20 13:04:35 -07:00
29089c7b41
Deprecation of qps_driver and use of shell scripts, in progress
2016-04-20 12:38:16 -07:00
87bac959ee
Eradicate Uniform, Deterministic, and Pareto interarrival
...
distributions since we don't use them and it's not sensible
to add them
2016-04-20 09:35:35 -07:00
e9a8d89dc2
fix formatting
2016-04-15 15:38:17 -07:00
33c161dffa
populate ScenarioResult.summary in JSON report
2016-04-15 14:53:54 -07:00
6321cec58e
fix formatting
2016-04-14 20:00:29 -07:00
453442eefb
fix formatting
2016-04-14 18:05:42 -07:00
969ffaf5c6
Enable JSON reports for qps drivers
2016-04-14 13:13:25 -07:00
f2ba7fe037
integrate ScenarioResult proto into qps driver
2016-04-14 13:13:25 -07:00
8f98e0b9dd
Add missing file
2016-04-11 23:39:12 -07:00
3ab2fe0094
Rollup of changes from the latest import
2016-04-11 20:11:18 -07:00
0bda0b3e55
Configure performance scenarios with json
...
Run those json configurations as part of presubmits
2016-04-06 10:14:44 -07:00
6169d5f7b0
Update copyrights
2016-03-31 07:46:18 -07:00
9533d042d4
Fix includes
2016-03-25 17:11:06 -07:00
f3d7638435
Merge pull request #5877 from vjpai/make_generic_payloads_great_again
...
Properly set server payload config in QPS tests
2016-03-25 16:26:39 -07:00
2e1903638e
Fix copyrights
2016-03-25 14:33:26 -07:00
f40df23eeb
Auto-changes
2016-03-25 13:38:14 -07:00
b830fb3a03
Merge branch 'master' into make_generic_payloads_great_again
2016-03-24 12:54:25 -07:00
a4942a012d
@jtattermusch correctly pointed out that we are not setting
...
payload config in our server config. This affect any generic server
tests that use anything other than 0-byte responses: essentially,
server-streaming or bidi throughput tests.
2016-03-21 12:44:02 -07:00
ca2886a05c
Made the code simpler to parse for humans
2016-03-21 12:04:18 -07:00
711766dc75
Ensure that no #includes are inside of a namespace.
2016-03-21 11:30:59 -07:00
847baf6fe0
clang-format and fix a typo caused by saving
2016-03-14 17:23:50 -07:00
63326281d7
1. Remove all deadlines from the RPCs and shutdown in this code.
...
These tests (especially unconstrained versions) can get very
backlogged and may take a while to finish. We sometimes flake waiting
for that. This is not hazardous (IMO), as the scripts that run these
tests already have timeouts to make sure that these don't truly go on
forever.
2. Make the time spent in the benchmark phase actually be
benchmark_seconds rather than benchmark_seconds-warmup_seconds
as it is currently.
2016-03-14 17:12:52 -07:00
2697f2141f
Merge branch 'master' into debug_qps_stream
2016-03-10 23:53:43 -08:00
baaf65546a
Fix formatting
2016-03-07 22:39:11 -08:00
710d58cfae
Disable profiling in qps_worker for now
2016-03-07 20:00:50 -08:00
2d9476898b
Revert "Ensure that no #includes are inside of a namespace."
2016-03-04 11:05:42 -08:00
2a8c280370
sanity
2016-03-02 20:52:34 -08:00
7d757ca29d
Ensure that no #includes are inside of a namespace.
2016-03-02 19:44:16 -08:00
c47c9497fa
clang-format
2016-02-26 14:39:59 -08:00
60a7e3336d
Finish clients fully before finishing servers
2016-02-26 10:30:17 -08:00
e7042b5ddf
Simplify Mark processing code
2016-02-25 20:28:05 -08:00
02eda41d21
Do stream operations in a more idiomatic way: Write, WritesDone, Read until fail, Finish
2016-02-25 19:43:57 -08:00
e2d39e08f9
Timer->UsageTimer consistently
2016-02-24 12:36:15 -08:00
86b258c462
Merge pull request #5388 from vjpai/bourne_again
...
Use traditional Bourne shell syntax for piping stdout and stderr
2016-02-24 10:29:22 -08:00
6483c32f7c
BSD platforms (such as Mac) are likely to have /bin/sh refer to traditional
...
Bourne shell syntax, not Bash syntax. Change "|&" to "2>&1 |" since that's the
traditional way to redirect stdout and stderr to a pipe
2016-02-24 09:29:12 -08:00
4ea4599a50
Merge github.com:grpc/grpc into naming-crisis
2016-02-23 17:13:38 -08:00
19fa540100
Fix copyrights
2016-02-23 08:19:39 -08:00
6541cabe9d
clang-fmt
2016-02-22 22:41:53 -08:00
732a875fe8
Add a sanity test for name aliasing
2016-02-22 15:59:19 -08:00
2d435b9178
Fix the slowdown factor
2016-02-19 00:42:08 -08:00
3b28872210
1. Adjust tsan/asan/msan slowdowns according to documentation
...
tsan documentation says 2-20x, so set it at 20x
asan documentation says 1.2-2.7x, so set it at 3x
msan documentation says 2-4x, so set it at 4x
This is now much less optimistic than before
2. Reactive tsan tests for qps_test
3. Set CPU load for qps_openloop_test
4. Divide qps_openloop_test Poisson rate by the slowdown factor of
the configuration
2016-02-19 00:28:28 -08:00
2815f184a3
Document variable names
2016-02-17 13:55:03 -08:00
8a2ff73640
Use more variables for better explanations
2016-02-17 09:21:10 -08:00
ed3cd423b6
Use three different load-factors and document load-factor variable
2016-02-17 08:49:03 -08:00
ee53dc5e08
Merge branch 'master' into loaded
2016-02-16 13:51:38 -08:00
e40aefb829
Fix the timeout
2016-02-13 01:25:30 -08:00
8ca0faaf52
use timeouts with proper fixture multipliers rather than raw,
...
since tsan slows things down a lot and may complicate issues such
as shutdown.
2016-02-13 00:25:45 -08:00
1a06082265
Add 70% load scenario for testing with open-loop
2016-02-12 17:34:08 -08:00
6583176337
clang-format
2016-02-12 15:59:03 -08:00
7914e62d48
Attack of the const police.
...
I also added some unrequested const's
2016-02-12 15:57:08 -08:00
761bcb4a18
Cleanup to make it work with clang build
2016-02-11 11:59:04 -08:00
914455782f
Sanity (clang-format and copyright)
2016-02-11 06:08:58 -08:00
1132c6b7db
Simplify and properly implement openloop testing.
...
Revive it in tests as well
2016-02-11 06:05:24 -08:00
9982c6fc3a
Fix random number generators and improve code style (stop using
...
operator overloads)
2016-02-10 14:55:33 -08:00
7a55684a13
Clang format and fix copyrights
2016-02-05 11:33:30 -08:00
bcfe99a547
Merge pull request #5008 from vjpai/worker_quit
...
Provide an RPC to quit the performance benchmarking workers
2016-02-04 16:32:03 -08:00
e4fdcd9242
Merge pull request #5013 from a-veitch/tag_set_to_context
...
Change all instances of census_tag_set to census_context
2016-02-04 12:42:35 -08:00
ed53c08238
Merge branch 'master' into worker_quit
2016-02-04 11:50:45 -08:00
9ded5dbf35
fix missed merge issue
2016-02-03 11:24:11 -08:00
7f1b65b8c6
Took out dependence on vector
2016-02-03 09:10:14 -08:00
574020edf1
Forgot to put this in the Mac version
2016-02-03 07:58:23 -08:00
ddb163a0e4
merge
2016-02-02 13:33:44 -08:00
2066443b7c
Function name changed on merge....
2016-02-02 11:48:12 -08:00
b47b556d33
Merge branch 'master' into worker_quit
2016-02-02 10:36:12 -08:00
2ed337aab3
clang-format
2016-02-02 09:43:35 -08:00
75d5c0f024
post merge
2016-02-02 09:43:02 -08:00
21e5d2b2f1
Add a Quit RPC so that we can conveniently shut down the workers from the driver.
2016-02-02 09:36:36 -08:00
daadcc8f74
at
2016-02-01 16:49:55 -08:00
33e51184fc
Address reviewer comments regarding const and shortage of comments
2016-02-01 16:40:06 -08:00
0d7a070e7e
Add core-limited scenario and protobuf-based QPS scenario
2016-02-01 14:10:47 -08:00
595674275b
Copyright and clang-format
2016-02-01 13:28:55 -08:00
6b05639ab1
Make sure that client limit is valid
2016-02-01 13:01:34 -08:00
vjpai