Network hickups to domU's when running a lot of domU's

Bug #376451 reported by Wido den Hollander
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
xen-meta (Ubuntu)
New
Undecided
Unassigned

Bug Description

Hello,

I have two systems with the following specifications:

- 2x Intel(R) Xeon(R) CPU E5420
- 64GB DDR2 FB-DIMM
- 2x Intel 80003ES2LAN
- SuperMicro X7DB8 mainboard
- Areca ARC-1680ix-16 RAID Controller

On these machine's in run Ubuntu 8.0.4.2 with Xen 3.2

The kernel has been customized to raise the number of NR_DYNIRQS, please refer to my other bug report (#341846)

Now i have tackled that problem but now i am starting to have issues with network connectivity for existing domU's. They run fine, but for no good reason they loose their network for about 2 minutes at the time while the dom0 works fine.

I have attached some data about the system (like xm list, xm dmesg, lspci, etc, etc)

Like i said, i have two systems, on with about 100 domU's (system A) and the other with 10 domU's (system B). They are both 100% identical in software and hardware, but only the domU's on the system with 100 domU's are experiencing these problems.

On system A i found the following messages when running "xm dmesg"

(XEN) printk: 271 messages suppressed.
(XEN) grant_table.c:1262:d0 Bad flags (0) or dom (0). (expected dom 0)

After some Googling i found old messages about a issue in the Xen grant table, but no real solution.

Now one of the interesting things is the messages i found inside a domU:

[135111.985303] netfront: rx->offset: 0, size: 4294967295
[135111.985330] netfront: rx->offset: 0, size: 4294967295
[135111.985337] netfront: rx->offset: 0, size: 4294967295

This indicates something goes wrong inside netfront.c, but for this i also found some reports, but no real solutions.

System B is not running a lot of domU's at the moment, but i think these problems will also rise on that system as soon as it hits some kind of barrier.

I hope somebody has a clue what is going on.

Tags: netfront xen
Revision history for this message
Wido den Hollander (wido) wrote :

I forgot to attach this file

tags: removed: e1000
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers