Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/16658 )

Change subject: KUDU-3210 Increase key size in tests and EMC
......................................................................


Patch Set 3:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/16658/3//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/16658/3//COMMIT_MSG@11
PS3, Line 11: This commit removes the test-only key size
> The master test, for example, has an assertion that would fail depending on
OK, then that's the only difference in 
https://gerrit.cloudera.org/c/16658/3/src/kudu/master/master-test.cc#1970 so 
far.  I guess developers adding new tests now and in the future don't need to 
worry about key sizes, unless that's something directly related to encryption 
and TLS.  If they touch something related to TLS and encryption, by definition 
they should know what they are doing and should care about key size, so I'm not 
buying your argument.

I think the verification that our tests pass in FIPS-approved mode doesn't 
require running these on pre-commit builds and elsewhere.  Running FIPS-enabled 
mode time to time and gating a new release on the results seems like a viable 
approach to me.

I think that unit tests and other tests we have in our test suite are targeted 
to verify specific functionality and features, and track regressions, if any.  
I don't think we have a need to run _all_ the suite with longer keys all the 
time.  If there a need to test some particular crypto-related aspects of using 
TLS and wire encryption in Kudu, let's figure out what we need to focus and add 
a targeted scenario that would run as a pre-commit verification.

To mirror real-life behavior we should run integration-level tests.  Unit tests 
are not exactly for that purpose, IMO.


http://gerrit.cloudera.org:8080/#/c/16658/3//COMMIT_MSG@11
PS3, Line 11: This commit removes the test-only key size
> In the meantime I ran the full test suite (without slow tests) serially on
Thank you for benchmarking those!  I guess perf statistics for ctest itself 
isn't that important in this context (rather it's important to see how each 
test affected since that's where the crypto ops are actually taking place), but 
time elapsed gives us good information on what's the overall increase in time 
for the test suite.

BTW, I also run the suite several times at CentOS 7.4 with OpenSSL 1.0.2k-fips  
26 Jan 2017, both in FIPS mode with the patch and in non-FIPS mode without the 
patch, in both cases with KUDU_ALLOW_SLOW_TESTS=1, RELEASE build configuration. 
 I ran them as 'ctest -j4' on a machine that has more than 4 CPU cores 
available.

I got the following results from the latest runs (I also confirmed that the 
variation from run to run was very small):

non-FIPS mode, without the patch:
  Total Test time (real) = 7707.15 sec

FIPS mode, with the patch:
  Total Test time (real) = 9319.28 sec

The difference is about 1.21 times increase in run-time, i.e. more than 20% 
increase from current.

With that, I'm not sure we want to increase runtime of the tests we run at 
every pre-commit check, even if we run those via dist-test.


http://gerrit.cloudera.org:8080/#/c/16658/3/src/kudu/master/master-test.cc
File src/kudu/master/master-test.cc:

http://gerrit.cloudera.org:8080/#/c/16658/3/src/kudu/master/master-test.cc@a1970
PS3, Line 1970:
Why don't to keep this comment explaining the semantics behind the magic number 
of the signature size?  Yes, it changes with the key size, but I think there is 
a simple correspondence between the RSA key size of the size of the signature 
if using SHA256 digest.



--
To view, visit http://gerrit.cloudera.org:8080/16658
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I45b96e0b4499ea1d80db5871a529c732ad41220e
Gerrit-Change-Number: 16658
Gerrit-PatchSet: 3
Gerrit-Owner: Attila Bukor <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Attila Bukor <[email protected]>
Gerrit-Reviewer: Grant Henke <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Tidy Bot (241)
Gerrit-Reviewer: Wenzhe Zhou <[email protected]>
Gerrit-Comment-Date: Mon, 02 Nov 2020 18:22:13 +0000
Gerrit-HasComments: Yes

Reply via email to