- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I am using X722 Ethernet Adapter on my server. System monitor reports that there are packets dropped on the network interface(reports around every 30 seconds), as I dig into the driver code (i40e), I find the error is rx_dropped. But my server still runs well. So I am wondering how to capture the dropped packets on the ethernet interface which cause rx_dropped to increase so that I can better find the possible bugs. Thank you.
Best
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
Thank you for patiently waiting for our update.
Upon checking with our engineering team, there is no way to monitor ring usage while the system is running.
And also, upon looking at details that you provided, this is not a retail X722-DA2.
The list shows X722 for 10GBASE-T, the only X722 cards are SFP+ not Twisted pair.
And the EtrackID shown does not match up with any cards that we have.
The NVM shows 3.25, there should be a newer version, depends on if it's a LOM and who the OEM is.
So for us to know where to direct you for further support. Kindly provide the following details.
- What is the model of your board?
- I know I already asked this before and you said that the card is already installed in the server but having the markings of the card should tell us what is the manufacturer of the card.
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello ,
Thank you for posting in Intel Ethernet Communities.
For us to understand the issue better, I need to gather the following details:
- Are you using onboard network card or a PCIe card?
- What is the brand and model of your server?
- Please provide a screenshot of the pocket loss test.
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
I just want to make a follow up on the information that I requested below.
- Are you using onboard network card or a PCIe card?
- What is the brand and model of your server?
- Please provide a screenshot of the pocket loss test.
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Michael,
Thanks for your reply!
- I am using a PCIe card. X772. Here is some output after I run ethtool --driver LAN and lspc
driver: i40e
version: 1.4.25-k
firmware-version: 3.25 0x800009d3 1.1638.0
expansion-rom-version:
bus-info: 0000:1a:00.1
supports-statistics: yes
supports-test: yes
supports-eeprom-access: yes
supports-register-dump: yes
supports-priv-flags: yes
And some hardware configuration.
- The server is customized server. And the cpu, RAM were not fully loaded (~70%) when drop happened.
- the attach file is a simple log recording the udp socket buffer usage, some result given by 'ethtool -S LAN', and 'ethtool -c LAN'(we turn on the adaptive rx). and the number of intterupts each network queue generate every 0.5s. The log is quite self-explainable. And the expected throughput is around 400Mb/s.
The drop happened on several servers, after we changed the ring buffer size from 512 to 2048, most of the server worked well but only one server still has packets dropped. These servers are of the same configuration. I am very confused. If you need more information, just let me know. Thank you!
Best,
Jimmy
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The previous reply may not be much relevant to the post title, but I am working on the packet lossing problem on the ethernet interface. They all help.
- Tags:
- The
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
Before we proceed, let me clarify the part when you said that the server is a customized server and the other components are not running on normal configuration. Are you designing a system?
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Michael,
I am not designing the system, other components are running on normal configuration. What I mean is that when the drops happened, other components (cpu,RAM) are running well. Attached is the full hardware component list. Thank you.
Best,
jimmy
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
Thank you for the clarification. Let me gather more details about the network card for me to further assist you.
- You mentioned that the you are using a PCIe X722, please provide the PBA number of the card of a picture of the card focusing on the markings.
- Where did you download the driver that you used?
- You also mentioned that you have other server with similar configuration and only one is having pocket drops. Have you tried updating the firmware of the X722 card?
- Can you also Run the command ethtool –i
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Michael,
1. The card is already installed on a remote server to which I can not have access right now. I am afraid I can not provide such information.
2. I am using the system default driver(i40e), here is the link to kernel https://github.com/alibaba/cloud-kernel/releases?after=v4.19.70
3. No, but all the servers are of the same configuration and running the same code. Increasing the ring buffer from 512 to 2048 solved the dropping problem for them but not for this one. That's why I get confused. ( Increasing the buffer size to 4096, reduces the drop but also introduces unacceptable delay to the our code).
4. ethool -i LAN gives :
driver: i40e
version: 2.3.2-k
firmware-version: 3.25 0x800009d3 1.1638.0
expansion-rom-version:
bus-info: 0000:1a:00.1
supports-statistics: yes
supports-test: yes
supports-eeprom-access: yes
supports-register-dump: yes
supports-priv-flags: yes
To sum up, what I want to know is: 1) how to monitor the ring buffer usage when the system is running. As binary search for the optimal ring buffer size costs too much. 2) how to capture the dropped packets( like, directly dumping them to the disk?). Thank you!
Best,
Jimmy
TuSimple, Inc.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
Thank you for providing the information that I requested. Let me further check your 2 questions for us to provide an answer.
Please give us 2 to 3 working days to provide an update.
Best regards,
Michael L.
Intel Customer Support Technicians
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello jimmy_tusim,
Thank you for patiently waiting for our update.
Upon checking with our engineering team, there is no way to monitor ring usage while the system is running.
And also, upon looking at details that you provided, this is not a retail X722-DA2.
The list shows X722 for 10GBASE-T, the only X722 cards are SFP+ not Twisted pair.
And the EtrackID shown does not match up with any cards that we have.
The NVM shows 3.25, there should be a newer version, depends on if it's a LOM and who the OEM is.
So for us to know where to direct you for further support. Kindly provide the following details.
- What is the model of your board?
- I know I already asked this before and you said that the card is already installed in the server but having the markings of the card should tell us what is the manufacturer of the card.
If you have questions, please let us know. In case we do not hear from you, we will make a follow up after 3 workings days. Thank you.
Best regards,
Michael L.
Intel Customer Support Technicians
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page