auth jwt: grpc context timed out for ErrAuthOldRevision

[xqzhang2015]

We're trying to build a C++ client for V3, with supporting TLS and auth. But when testing auth error ErrAuthOldRevision, client-side doesn't receive any expected response and just grpc context timed out. Not sure it's an etcd cluster issue or grpc c++ issue.

You could reproduce it by sync or async get call. Steps:

1. authenticate and call get
2. add a new user(with it, auth revision will be increased with 1)
3. call get again(Expectation: ErrAuthOldRevision response, Physical: grpc context timed out)

[etcd --version]
etcd Version: 3.3.13
Git SHA: 98d3084
Go Version: go1.10.8
Go OS/Arch: linux/amd64

[grpc version]
grpc C++, 1.14.0

[System]
Linux 3.10.0-229.el7.x86_64 #1 SMP Fri Mar 6 11:36:42 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

[mitake]

@xqzhang2015 can you reproduce this on the latest master branch? Probably it would be caused by the invalid version number mechanism implemented in the server side (see also #10408).

[xqzhang2015]

@mitake thanks for the info. And it's really caused by the infinite loop issue of JWT token revision staleness.

Once this client timed out issue happened, the CPU utilization of etcd cluster will become super high, just as same as the #10408 or #10218 .

With that patch for v3.3.13, the client timed out caused by JWT token staleness disappeared.

I noticed that v3.4.0-rc.1 has included this patch, but #10468 is not included, will it be applied? BTW, will the official v3.4.0, instead of experimental version, be released on 2019-08-30?

Furthermore, will it be included by v3.3.14? If yes, when will the official version be released?

Thanks again. Really appreciate your confirmation.

[xqzhang2015]

@mitake or @xiang90 would you like to give any comment for my question above?

[mitake]

@xqzhang2015 sorry for my late reply. @gyuho how do you think about backporting the change? If it's ok for you, I'll open a PR for backporting.

[xqzhang2015]

I noticed v3.3.15 was released recently and think it's ok for me to release it to next v3.3.x version.

[mitake]

@xqzhang2015 got it, let me handle this.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 21 days if no further activity occurs. Thank you for your contributions.