On May 2, 2019, at 11:30 AM, Robert Varga <nite@...> wrote:



On 02/05/2019 20:05, Luis Gomez wrote:

Hi netconf devs,

A CSIT regression has been detected in Neon since 2 days:

https://jenkins.opendaylight.org/releng/job/netconf-csit-1node-userfeatures-all-neon/

Connected to this, there is this BC exception in the testtool log:

08:44:28.082 [main] DEBUG o.a.s.c.util.security.SecurityUtils - register(BC) not registered - enabled=false, supported=true
08:44:28.083 [main] DEBUG o.a.s.c.util.security.SecurityUtils - register(EdDSA) not registered - enabled=true, supported=false
08:44:28.086 [main] ERROR o.o.netconf.test.tool.Main - Unhandled exception
java.lang.IllegalArgumentException: BouncyCastle not registered
at org.apache.sshd.common.util.ValidateUtils.createFormattedException(ValidateUtils.java:213)
at org.apache.sshd.common.util.ValidateUtils.throwIllegalArgumentException(ValidateUtils.java:179)
at org.apache.sshd.common.util.ValidateUtils.checkTrue(ValidateUtils.java:156)
at org.apache.sshd.common.util.security.SecurityUtils.createGeneratorHostKeyProvider(SecurityUtils.java:503)
at org.opendaylight.netconf.test.tool.NetconfDeviceSimulator.getPemGeneratorHostKeyProvider(NetconfDeviceSimulator.java:294)
at org.opendaylight.netconf.test.tool.NetconfDeviceSimulator.start(NetconfDeviceSimulator.java:194)
at org.opendaylight.netconf.test.tool.Main.main(Main.java:47)

So when I change the test tool command like in this patch, then testtool works fine:

https://git.opendaylight.org/gerrit/#/c/81879

Thanks for finding that.

Since I am not sure why this BC option was introduced in first place, my question is: should we go ahead and merge the test patch or is there anything else we need to investigate?

https://git.opendaylight.org/gerrit/#/c/32070/ provides a clue, esp.
pointer to
https://wiki.opendaylight.org/view/OpenDaylight_Controller:Netconf:Testtool#Slow_creation_of_devices_on_virtual_machines

Given that refers to VMs, my guess it has to do with RNG and seeding --
not sure if it still applies.

In general we have stopped pretending like we do not use BC and embrace
it, so disabling should be the option of last resort. It would seem
sshd-core started requiring it for some operations, even ...

Regards,
Robert