Hi,

From the bug description and screenshot, the issue happens once after the system was installed on a bare metal server (Simplex). ovs-vswitchd processes that you suspected are running in background when 1^st time the system was booting up (so more like a memory used up rather than memory leak). Correct?

I’d like to check whether and how OVS supports multiple ovs-vswitchd processes (which seems abnormal to me). Would you please kindly provide the following details which may help with the triage?

1. kernel version and OVS version that you’re using;

2. OVS or OVS-DPDK;

3. full screenshot (with full cmds) and additional OVS logs if possible.

BTW,

1. can manually killing all OVS processes followed by starting/configuring an OVS instance/daemon work around the issue?

2. does anyone have a remote debug environment so that we could better look into the issue?

Thanks!

BR,

Kailun

From: Xie, Cindy [mailto:cindy.xie@intel.com]
Sent: Friday, August 31, 2018 5:47 PM
To: starlingx-discuss@lists.starlingx.io; li, <kai@99cloud.net>
Subject: [Starlingx-discuss] mem run our in StarlingX Simplex bare metal install

All,

99Cloud reported that they encountered mem run out issue after StarlingX being installed in their Simplex bare metal. Below are some info provided by them:

HW config:

- CPU: 56c

- Memory: 188GB

- Hard Disck: 2400G SAS RAID1*1/480G SSD RAID1*1

- OS: bootimage0727.iso

- NIC (see below picture)

Issue: once the system being installed on bare metal server (Simplex), check the mem usage, they see the issue below:

They didn’t have chance to create any tenant VM yet due to mem is almost fully occupied (170GB).

@Li Kai, please add additional info if the above is not complete.

Thanks. - cindy