Project

General

Profile

Support #4021

long-term end-to-end osmo-remsim test setup

Added by laforge 7 months ago. Updated 8 minutes ago.

Status:
In Progress
Priority:
High
Assignee:
Category:
tests
Target version:
-
Start date:
05/23/2019
Due date:
% Done:

50%


Description

let's build a setup consisting of:
  • 5-slot CCID card reader
  • sysmoQMOD with four Gobi2000 modems
    • four random SIM cards
  • APU2 with Debian9
  • one instance of remsim-server from osmocom:nightly feed running on the host
  • Debian 9 VM for remsim-bankd
    • USB port towards card reader routed to this VM
    • one instance of osmo-bankd from osmocom:nightly feed running
  • Debian 9 VM for remsim-client
    • USB port towards QMOD routed to this VM
    • four instances of osmo-remsmim from osmocom:nightly feed running

The host machine should run tc netem to simulate a non-zero latency, such as 200ms between client/server/bankd

All software should automatically start up after boot.

History

#1 Updated by roh 7 months ago

  • assembled hardware
  • configured host system for libvirt, remsim-server, sshd
  • moved to osmocom-network segment (vlan25)
  • added cardreader

#2 Updated by roh 7 months ago

done:
  • 2x kvm debian9 inside the debian9 on apu2 for remsimbank and remsimclient(s)
  • added nightly feeds to all debian
  • usb from acs reader routed into remsimbank vm
  • autostart vms
  • installed osmo-* daemons in respective vm
  • network segment for vms: 10.9.99.0/24 - .1 server .10 client .30 simbank
todo:
  • tc netem
  • systemd servicefiles for all daemons
  • qmod/modems
  • simcards

#3 Updated by roh 7 months ago

done:
  • systemd servicefiles for -bankd and -server
  • qmod/modems connected - may need to move to a different usb controller (issues pushing all parts of the device to vm)

#4 Updated by roh 7 months ago

done:
  • tested usb forwarding via complete seperate usb port - no go on apu due to hw limits.
  • swapped the vm for -client and host for -server due to usb/pci forwarding limitations (qmod needs to be on host)
  • systemd servicefile template (copy 4 times and configure for specific modems)

#5 Updated by roh 6 months ago

done:
  • systemd servicefile for osmo-remsim-client using proper templating (no copying servicefile around anymore), just add more configs.
todo:
  • netem
  • manual testing/verification

#6 Updated by roh 5 months ago

done:
  • updated all install/vm
  • the clients now have proper ids (-n 0-3)
  • basic netem setup (inmarsat-style 300ms / 150ms jitter / 50% corrlation / 512kBit)
    on network output of remsimserver-vm - do we need a more complicated setup here? it feels proper laggy to me.
todo:
  • real tests/load/traffic on all components (do we have a list of things to probe already?)
  • which testrunner to use?

#7 Updated by roh 5 months ago

#8 Updated by laforge 5 months ago

  • Priority changed from Normal to High

#9 Updated by roh 3 months ago

i think we need a bigger ssd for this machine. it ran out of disk twice now.

it has 2 images for the vm (4g each) and 4g swap, which results in about 500mb free when i delete all old logs.
i tried adding it to the monitoring but we have another blocking issue there. (#4664)

kevin got access a few days ago, so he can help get some traffic on those simcards.

how will this work exactly? jenkins will log in via ssh and key and trigger some script?

#10 Updated by roh 3 months ago

done:
  • swapped the simcards to sysmoUSIM.

#11 Updated by roh about 2 months ago

done: * upgraded ssd to 240G, copied system, fixed uuid, resized / to 100Gb, grown swap to 16Gb, added remaining diskspace to /space

#12 Updated by laforge 10 days ago

I cannot find any mention of the IP adresss of thee system. roh please add.

#13 Updated by roh 10 days ago

its on 10.9.25.90/24

i noted it in https://osmocom.org/projects/osmo-remsim/wiki/Osmo-remsim#Test-Suite (only on the graphic)

#14 Updated by laforge 8 minutes ago

  • Status changed from New to In Progress
  • Assignee changed from roh to laforge
  • % Done changed from 0 to 50

I'm currently upgrading the system and the VMs to buster. We shouldn't start to deploy new systems based on oldstable, it just means we need to migrate at some point sooner than neccessary.

Also available in: Atom PDF

Add picture from clipboard (Maximum size: 48.8 MB)