Difference between revisions of "CSIT/CSIT LF VIRL testbed"

Revision as of 10:35, 16 May 2017

VIRL infrastructure open tasks

This is the current working list of identified tasks for CSIT VIRL testbeds. It is updated periodically.

High Priority Tasks

1. [WIP] Detecting and clearing stuck VIRL simulations
   1. Description: Continue getting stuck VIRL simulations due to either LF network connectivity interruptions or failing CSIT bootstrap teardown.
   2. Solution: automate clearing old (garbage) simulations, detect non-successful simulation teardowns.
   3. Tasks:
      • [IN-REVIEW] Use built-in VIRL simulation expire timer set to 2hrs.
        • coded default simulation expiry to 120min, semi-weekly to 500min, weekly to 120min. https://gerrit.fd.io/r/#/c/6656/.
      • [OPEN] Extend CSIT bootstrap teardown stop-simulation API call to verify if it was SUCCESS/FAIL.
2. [WIP] Lack of VIRL server healthcheck in CSIT
   1. Description: no regular automated healthchecks executed against VIRL servers.
   2. Solution: introduce a CSIT health-check monitoring job for VIRL servers' health.
   3. Tasks:
      • [OPEN] Create a new job, executed periodically (6hrs?) for healthchecking all VIRL servers.
      • [OPEN] VIRL health-check APIs: health status, VIRL API tests, simulation tests.
      • [OPEN] VIRL check for old (garbage simulations).
      • [IN-REVIEW] pre-check to every start-testcase to better handle exceptions and printing errors. https://gerrit.fd.io/r/#/c/6656/.
3. [OPEN] VIRL simulation IPv4 address depletion
   1. Description: Today there is one /24 subnet allocated for all VIRL simulations, split equally across 3 servers, 84 /32 addresses per server. Each CSIT simulation takes 4 addresses (mgmt, tg, sut1, sut2), each csit-vpp and vpp-csit verify job uses 3 simulation to parallized tests for reduced execution time. This means each server has capacity to run up to 7 verify jobs concurrently (3*4*7). Once Centos7 tests productized, where two jobs are always executed in parallel, this will reduce it down to 3 concurrent jobs. Not good. It's basically a show stopper to productize Centos7 into vpp-csit-verify per patch jobs.
   2. Solution options: Need to increase IPv4 address space given to VIRL hosts. Dedicating /24 subnet per VIRL server, will give address capacity for 60 concurrent simulations. Based on previous memory calcs each VIRL host is capable of doing 30 simulations (30*3 VMs) - need to check this.
   3. Tasks:
      • [OPEN] Check the current max CSIT simulation capacity per VIRL host. Test it.
      • [WIP] Request LF FD.io infra team to allocate at least another /24 for VIRL simulations.
        • [WIP] Request acknowledged by LF as part of [FD.io Helpdesk #40733]. https://lists.fd.io/pipermail/csit-dev/2017-May/001911.html.
4. [OPEN] Script expecting VIRL sim nodes to be active within ca. 120sec after launch request - this is too tight
   1. Description: Intermittent test job failures due to 'ERROR: Simulation started OK but devices never changed to ACTIVE state’. Number of these can be avoided by increasing the script timeout to 240sec or so.
   2. Solution: Increasing the script timeout to 240sec or so. But don’t wait 4min every time before trying, as this will add to the overall execution time.
   3. Tasks:
      • [OPEN] Increase test script timeout to 240sec(?).
5. [WIP] tb4-virl1 server upgrade to ubuntu16.04, VIRL-core ver. 10.32.8, OpenStack Mitaka.
   1. Description: virl upgrade to address issues with Centos7 test instabilities related to QEMU, and to improve general virl system robustness.
   2. Solution: upgrade tb4-virl1 server to ubuntu16.04, VIRL-core ver. 10.32.8, OpenStack Mitaka. verify stability. follow gradually with tb4-virl1 and then tb4-virl2 upgrades.
   3.  Tasks:
      • [WIP] VIRL1 server 10.30.51.28 - currently in STAGING, resolving issues.
      • [OPEN] VIRL1 server 10.30.51.28 - move to PRODUCTION once determined stable. Monitor PRODUCTION performance.
      • [OPEN] VIRL2 server 10.30.51.29 - upgrade, verify stability, move to PRODUCTION.
      • [OPEN] VIRL2 server 10.30.51.30 - upgrade, verify stability, move to PRODUCTION.
6. [OPEN] Need to periodically delete old files in /tmp directory
   1.  Tasks:
      • Cron job to delete old (more then 2 weeks?) files in /tmp directory on every VIRL server

        crontab -e

        0 0 * * * * find /var/log/libvirt/qemu -type f -mtime +14 -name "instance*.log" -delete

        0 0 * * * * find /tmp -type f -atime +14 -name "*.deb" -delete

        0 0 * * * * find /tmp -type f -atime +14 -name "*.rpm" -delete

        0 0 * * * * find /nfs/scratch/ -type d -mtime +14 -name "session-*"

7. [OPEN] - Currently the latest nested VM image is used for all Ubuntu/Centos images
   1. Description: need solution to be able to link different nested VM images to different ubuntu/centos images
8. [WIP] Centos7 tests productization
   1. Description: Following upgrade of tb4-virl1, Centos7 tests should be ready for productization.
   2. Solution: Proposal to run Centos7 tests periodically (daily) instead of per patch, to avoid VIRL simulations overload.
   3. Tasks:
      • [OPEN] Create a daily vpp-csit-verify-Centos7 job.

Other Tasks

1. CSIT-116 [1]: Modify VIRL and nested-VM username/password
2. CSIT-159 [2]: Nested VM: Replace cisco/cisco credentials with csit/csit
3. CSIT-160 [3]: Ubuntu VM: Replace cisco login with csit
4. CSIT-145 [4]: Out-of-band access to SUTs
5. CSIT-151 [5]: Do not destroy VM in case of test failure due to infrastructure issue
6. CSIT-150 [6]: Health-check to capture TG/SUT environment after failed test case
7. CSIT-202 [7]: Execute start/stop-testcase scripts from git repository
8. CSIT-115 [8]: Usage and status monitoring of VIRL hosts
9. CSIT-112 [9]: VIRL infrastructure periodic creation and distribution of images
10. CSIT-90 [10]: Nested-VM boot-up failed
11. CSIT-210 [11]: Nested VM to include l3fwd startup script
12. CSIT-161 [12]: Update nested VM qemu library to use 3rd serial console
13. CSIT-356 [13]: Update VIRL testbed creation to allow specification of centos image