Hi Paul.

> "cli_inband" commands

There are other ways how CSIT can be broken by an innocent looking VPP change.
CSIT currenlty does use:
- VAT console commands
- vppctl commands
- cli_inband via PAPI
- in-house serialization-deserialization for using PAPI remotely [1]

Other VPP interfaces which affect CSIT if changed:
- startup.conf (and defaults there)
- initialization of vpp_papi.py
- structure of vpp stats

Some of the risks are being addressed,
(for example I am working on how to avoid
the need of CSIT Python utils on VPP machine side),
but some risks will stay for a while.

For example, in order to read vpp stats from remote machine
(without the above-mentioned python utils and custom serialization),
vpp_get_stats app [2] accessed over SSH is a possibility,
but having an equivalent CLI call will be an improvement.

> I am against this.
> but I am already aware of them.

I agree that CSIT should be moving towards officially supported APIs.
But a job able to spot a danger in an innocent looking VPP Change
is even more important before CSIT gets there.

Vratko.

[1] 
https://gerrit.fd.io/r/#/c/18360/60/resources/tools/papi/vpp_papi_provider.py
[2] https://github.com/FDio/vpp/blob/master/src/vpp/app/vpp_get_stats.c

From: vpp-dev@lists.fd.io <vpp-dev@lists.fd.io> On Behalf Of Paul Vinciguerra
Sent: Friday, 2019-May-24 15:31
To: vpp-dev@lists.fd.io
Subject: Re: [vpp-dev] device job voting?

Hi Vratko,

As long as CSIT uses "cli_inband" commands, I am against this. (This practice, 
by the way also short-circuits any/all flag day type efforts/proposals!)
Traceback (most recent call last): File 
"/w/workspace/vpp-csit-verify-device-master-1n-skx/csit/resources/libraries/python/VPPUtil.py",
 line 149, in verify_vpp_on_all_duts VPPUtil.verify_vpp(node) File 
"/w/workspace/vpp-csit-verify-device-master-1n-skx/csit/resources/libraries/python/VPPUtil.py",
 line 136, in verify_vpp VPPUtil.show_log(node) File 
"/w/workspace/vpp-csit-verify-device-master-1n-skx/csit/resources/libraries/python/VPPUtil.py",
 line 290, in show_log return papi_exec.add('cli_inband', cmd='show 
log').get_replies()

Can we compile a list of all the missing API's somewhere (the wiki maybe?) 
where "cli_inband" needs to be used to get the job done?

Btw, this is not just a CSIT issue.  This is also an issue with the python 
tests and the vpp_pg_interface code as well, but I am already aware of them.

Paul
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#13141): https://lists.fd.io/g/vpp-dev/message/13141
Mute This Topic: https://lists.fd.io/mt/31639207/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
  • ... Vratko Polak -X (vrpolak - PANTHEON TECHNOLOGIES at Cisco) via Lists.Fd.Io
    • ... Paul Vinciguerra
      • ... Vratko Polak -X (vrpolak - PANTHEON TECHNOLOGIES at Cisco) via Lists.Fd.Io
      • ... Vratko Polak -X (vrpolak - PANTHEON TECHNOLOGIES at Cisco) via Lists.Fd.Io

Reply via email to