dokucama

User Tools

Site Tools

Trace:
network_stuff:infiniband

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
network_stuff:infiniband [2022/05/20 11:48] – jotasandokunetwork_stuff:infiniband [2025/10/03 16:04] (current) – jotasandoku
Line 1: Line 1:
 __**INFINIBAND**__: __**INFINIBAND**__:
 +\\
 +{{ :network_stuff:infiniband_technical_guide_for_network_engineers.pdf |}}
 \\ \\
 From [[https://en.wikipedia.org/wiki/InfiniBand|wikipedia]]: InfiniBand (IB) is a computer networking communications standard used in high-performance computing that features very high throughput and very low latency.[..]. InfiniBand provides remote direct memory access (RDMA) capabilities for low CPU overhead. More info [[https://network.nvidia.com/related-docs/whitepapers/InfiniBandFAQ_FQ_100.pd|here]]  From [[https://en.wikipedia.org/wiki/InfiniBand|wikipedia]]: InfiniBand (IB) is a computer networking communications standard used in high-performance computing that features very high throughput and very low latency.[..]. InfiniBand provides remote direct memory access (RDMA) capabilities for low CPU overhead. More info [[https://network.nvidia.com/related-docs/whitepapers/InfiniBandFAQ_FQ_100.pd|here]] 
Line 10: Line 12:
   * RDMA provides access to the memory from one computer to the memory of another computer without involving either computer’s operating system. This technology enables high-throughput and low-latency networking with low CPU utilization.   * RDMA provides access to the memory from one computer to the memory of another computer without involving either computer’s operating system. This technology enables high-throughput and low-latency networking with low CPU utilization.
     * Mellanox provides RDMA via the OFED package     * Mellanox provides RDMA via the OFED package
-  * lid : local indentifier (All devices in a subnet have a Local Identifier (LID)). Routing between different subnets is done on the basis of a Global Identifier (GID) +  * **LID** : local indentifier (All devices in a subnet have a Local Identifier (LID)). Routing between different subnets is done on the basis of a **Global Identifier (GID)** 
-  * NSD (Network Shared Disks): In our context, NSD is the server that connects to the storage via the Mellanox switch. The servers share the NSD's to the clients, creating some sort of distributed logical disk (a bit like the hyperflex technology). +  * GID: Is another identifier but is to route BETWEEN SUBNETS. Contains : Subnet Prefix and a GUID (Global Unique Identifier). 
-  * SM (Subnet Manager):  It performs the InfiniBand specification's required tasks for initializing InfiniBand hardware. One SM must be running for each InfiniBand subnet. It's run by the OpenSM daemon which can run bith in the switches and  the servers+  * NSD (Network Shared Disks): In our context, NSD is the server that connects to the storage via the Mellanox switch. The servers share the NSD's to the clients, creating some sort of distributed logical disk (a bit like the hyperflex technology). Particuartly in our setupm the servers dont share their local disks but they expose the DDN's disks. 
 +  * **SM (Subnet Manager):**  It performs the InfiniBand specification's required tasks for initializing InfiniBand hardware. One SM must be running for each InfiniBand subnet. It's run by the OpenSM daemon which can run bith in the switches and  the servers
     * SM master is the node truly acting as SM. The node with the highest priority [0-15] wins.     * SM master is the node truly acting as SM. The node with the highest priority [0-15] wins.
     * In our setup, servers all have priority 14 while switch has priority 15.     * In our setup, servers all have priority 14 while switch has priority 15.
Line 29: Line 32:
   show guids  ! To see the switch group identifier (like the switch main mac address)   show guids  ! To see the switch group identifier (like the switch main mac address)
   fae sminfo  ! To show who is acting as sm master. Note it can be a server or the switch itself. If the latter, 'show guids; and 'fae sminfo' thrown the same value   fae sminfo  ! To show who is acting as sm master. Note it can be a server or the switch itself. If the latter, 'show guids; and 'fae sminfo' thrown the same value
 +  fae ibnetdiscover   ! Show everything connected to the switches (servers and storage controllers) in a verbose format
      
 Useful server side ib commands Useful server side ib commands
-  ibstatus  # server infiniband (ib) interfaces status +  ibstatus  # server infiniband (ib) interfaces status + HCA model 
- +  ibping -S -L 10  ---- ibping -L 20 -c 10 -n 3     # to ping, we need to run one in the server (LID 20) and one in the client (LID 10). This is because even ping makes RDMA calls
- +
----- +
-Terms:\\ +
  
  
Line 47: Line 47:
  
 Tshoot commands: Tshoot commands:
-  show ib smnodes+  show ib smnodes    # also smquery can be useful for information about the SM itself.
   show ib smnode nyzsfsll51 sm-state   show ib smnode nyzsfsll51 sm-state
   show guids   # so we can identify the macs   show guids   # so we can identify the macs
Line 68: Line 68:
  
 ---- ----
-UPGRADE PROCEDURE:\\ 
-[[https://community.mellanox.com/s/article/howto-upgrade-switch-os-software-on-mellanox-switch-systems]] 
-(The device has two partitions. We can install the new OS in the 'backup' partition and instruct the device to boot from it).\\ 
-If not really a backup one, is the 'Active Image, partition of next boot', we set it up with ''image boot next'' 
  
-  * Configuration backup.  +MELLANOX UPGRADE 
-    * Better via UI:  Setup → Configuration → Configuration files (Active configuration file + Binary configuration file) +First scp the image to any of the linux servers. Preferably in the same region where the switch is. 
-      * If we needed to restore this configuration, we do: +\\ 
-  configuration fetch <download url> +Then do the following om the switch (this is an example): 
-  configuration switch-to <filename> # this is for binary configuration +  conf t 
-  configuration text fetch <download url> +  image delete XXX // --> delete old images, if exist 
-  configuration text file ss apply verbose fail-continue  # this is for the text configuration +  image fetch scp://myusername:mypassword@servername/lxhome/santosja/firmware/image-X86_64-3.6.6162.img 
-   +  image install image-X86_64-3.6.6162.img
-  enable  +
-  image fetch scp://[username:password@IP/image] +
-  image install im...+
   image boot next   image boot next
 +  configuration write
   reload   reload
-   + 
-  +The upgrade itself (after the reload) takes 3-4 minutes. 
 + 
 + 
 + 
 +To downgrade [[https://docs.mellanox.com/display/MLNXOSv391906/Downgrading+OS+Software]] 
  
 ---- ----
 +===== Simple InfiniBand Troubleshooting Case (1) =====
 +
 +**Symptom:**  
 +MPI (Message Passing Interface) jobs between two GPU nodes are limited to ~12 Gb/s, while other nodes on the HDR100 fabric reach 100 Gb/s.
 +
 +**Checks:**  
 +  * `ibstat` on the slow node shows:  
 +
 +    Port 1: State: Active  
 +            Physical state: LinkUp  
 +            Rate: 25 Gb/sec  
 +            Width: 1X
 +
 +    → Link has fallen back to 1 lane × 25 Gb/s instead of 4 lanes × 25 Gb/s.
 +
 +  * `iblinkinfo -r` confirms only **lane 1** is active and reports symbol errors on that port.
 +
 +  * `perfquery -x` shows increasing **SymbolErrorCounter** and **LinkErrorRecoveryCounter** on the affected switch port.
 +
 +**Fix:**  
 +Replaced the QSFP56 cable. `ibstat` now reports **Rate 100 Gb/s, Width 4X**. Bandwidth test (`ib_write_bw`) confirms ~97 Gb/s performance.
 +
 +**Root Cause:**  
 +One faulty cable lane caused fallback to 1X. Basic InfiniBand tools helped quickly identify and resolve the issue.
 +
 +
 +===== Simple InfiniBand Troubleshooting Case (2) Subnet Manager Misconfiguration =====
 +
 +**Symptom:**  
 +Some nodes fail to establish MPI communication or exhibit long startup times. `ibstat` shows ports stuck in "Initializing" or "Down".
 +
 +**Checks:**  
 +  * On affected nodes, `ibstat` shows:  
 +    Port 1: State: Down  
 +            Physical state: Polling  
 +            Rate: 100 Gb/sec (Expected)  
 +    → Port sees light but is not becoming Active.
 +
 +  * `ibv_devinfo` shows device present but no active port.
 +
 +  * No errors found in hardware or cabling.
 +
 +  * On a working node, `sminfo` reports a different Subnet Manager than the one expected:
 +
 +  SM lid 1, lmc 0, smsl 0, priority 5, state: master
 +  smguid 0x... "server-X"
 +
 +**Fix:**  
 +Two Subnet Managers (SMs) were active with equal priority, causing instability. Disabled one SM (`opensm`) on the unintended node. After restarting `opensm` on the correct node, all ports transitioned to "Active".
  
-Downgrade software: +**Root Cause:**   
-[[https://docs.mellanox.com/display/MLNXOSv391906/Downgrading+OS+Software]]+Multiple Subnet Managers with conflicting roles caused port initialisation to stall or flap. Ensuring a single master SM with correct priority resolved the issue.
  
  
network_stuff/infiniband.1653047329.txt.gz · Last modified: (external edit)

Page Tools

Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 4.0 International
CC Attribution-Share Alike 4.0 International Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki