chen

Member

Last active 5 days ago

  1. last week
    Thu Dec 7 07:53:25 2017

    Very cool!
    I assume you are speaking of ImageClassification. Is this correct?

  2. 3 weeks ago
    Thu Nov 23 07:13:27 2017
    chen posted in ComputerLog API.

    This can somewhat can simplify my code, for now I put a TODO comment to use the new API, for when we do the upgrade. Thanks.

  3. 5 weeks ago
    Mon Nov 6 07:06:43 2017
    chen posted in Logging level.

    We have a testing lab, that constantly deploy an image on a machine, runs the test, and do it again for a new test with a new image. We are using an API to drive this process (still not the official one that came with 1.3.0, that one is in the works, but one we hacked together).

    I got a report that a nightly test session failed because a scheduled deploy did not start. But after a manual reboot of the machine it did start.

    I wanted to gather data of what the system was asked to do, and what it was doing, to try and debug this issue.

  4. Sun Nov 5 16:28:27 2017
    chen started the conversation Logging level.

    Is there a way to increase the logging verbosity of components of C/D. I was able to do it for the tftpd server (by adding -v -v -v to the TFTP_OPTIONS at /etc/defaults/tftpd-hpa), but I don't know if I can rase the logging level of C/D DHCP Proxy, or the application running under apache2.

    Any advice?

  5. 6 weeks ago
    Mon Oct 30 09:24:37 2017
    chen started the conversation ComputerLog API.

    In the API we hacked together for C/D 1.2.1, by quiring the underlying database, we had a way (although not a pretty one) to get a log entry for a deployment job:

    Before we started a deploy job J for a ComputerId C, we found out the largest LogId L. Then the deploy job associated with J is the smallest LogId that is still larger than L where ComputerId is C.

    AFAIK, this surprisingly non-trivial task became even harder with the new API introduced by C/D 1.3.0:
    I can't find a good way getting the logs for a given computer. The best I can do is GET the /ComputerLog/GetOnDemandLogs, pick a sensible limit, filter the results, and if they do not contain the log entries I am looking for, increase the limit and repeat.

    Is there a better way for getting the log of a given deploy task?

  6. 7 weeks ago
    Wed Oct 25 15:57:22 2017

    That fixed it. Thanks.

    Will post in this thread in case I will find similar behavior elsewhere.

  7. Tue Oct 24 13:15:51 2017

    Just for posterity: I tried to stop apache2, clear /tmp and restart apache2 -- got the same results.

    I didn't know what logs should I look at, but if you need any other information I will do my best to post it here.

    Thanks.

  8. Tue Oct 24 13:06:32 2017

    The API we wrote to deal with this (and to work with v1.2.1), simply manipulates the allow list file, and restarts the cd_proxyd service. I guess there is no such stock API for C/D, and so we will keep this part of our code.

    May I suggest having such an API bundled with cd_proxyd?

  9. Tue Oct 24 13:00:15 2017

    Yes it crashes also with the fake computer entry.

  10. Tue Oct 24 12:54:50 2017

    I am using the stock mono that comes with Ubuntu 16.04

    $ mono --version
    Mono JIT compiler version 5.4.0.201 (tarball Fri Oct 13 21:51:45 UTC 2017)
    Copyright (C) 2002-2014 Novell, Inc, Xamarin Inc and Contributors. www.mono-project.com
            TLS:           __thread
            SIGSEGV:       altstack
            Notifications: epoll
            Architecture:  amd64
            Disabled:      none
            Misc:          softdebug 
            LLVM:          supported, not enabled.
            GC:            sgen (concurrent by default)

    All my hosts were defined in v1.2.1 (and I ported the database over to v1.3.0)
    I also have the old v1.2.1 running (the new host is a clone of the old one, upgraded) and in the old one I don't see this error.

    Should I delete my old computer entry and re-create it (with the same data)?

View more