[EXTERNAL] Re: Windows rte_pktmbuf_pool_create() ENOMEM

[Dmitry Malloy]

In order for us to raise this with proper team at MSFT, we need better diagnosis that "likely". 

Robert - can you enable any existing DPDK instrumentation/traces, or add more custom tracing for your setup to narrow down the root-cause?

Thank You,
Dmitry

-----Original Message-----
From: Dmitry Kozlyuk <dmitry.kozliuk at gmail.com> 
Sent: Wednesday, December 7, 2022 7:50 AM
To: Robert Hable <Robert.Hable at spusu.at>
Cc: users at dpdk.org; Tyler Retzlaff <roretzla at linux.microsoft.com>; Dmitry Malloy <dmitrym at microsoft.com>; Narcisa Ana Maria Vasile <Narcisa.Vasile at microsoft.com>
Subject: [EXTERNAL] Re: Windows rte_pktmbuf_pool_create() ENOMEM

Hi Robert,

2022-12-07 15:19 (UTC+0000), Robert Hable:
> after a few days of my windows machine running and starting/stopping my DPDK application multiple times, rte_pktmbuf_pool_create() fails with rte_error = EMOM.
> When closing most of the running programs on that machine, it usually works again for some time. I assume this happens because of memory fragmentation?
> What can I do against this error, except just rebooting the whole machine every time and closing applications all the time?

Yes, fragmentation is likely the cause.
Adding Microsoft people who might know how to prevent this, since I don't.