Re: hca_self_test.ofed: Firmware Check on CA #0 (VPI) .......... FAIL:REASON:...
Hi, Thank you for your reply ! I have tried the "ibv_devinfo", and I got the following information: run@gpu-cluster-1:~$ ibv_devinfo hca_id: mlx4_0 transport: InfiniBand (0) fw_ver: 2.11.500...
View ArticleRe: hca_self_test.ofed: Firmware Check on CA #0 (VPI) .......... FAIL:REASON:...
Could you also try to query the HCA with MFT package ? Mellanox Technologiesmst startmst statusflint -d *device id* q Reason I ask is that something is wrong with the discovery of your HCA's unique...
View ArticleRe: hca_self_test.ofed: Firmware Check on CA #0 (VPI) .......... FAIL:REASON:...
root@gpu-cluster-1:/home/run# mst status MST modules:
View ArticleClik here to view.
Re: hca_self_test.ofed: Firmware Check on CA #0 (VPI) .......... FAIL:REASON:...
Hello,root@gpu-cluster-1:/home/run# mst statusMST modules:------------ MST PCI module loaded MST PCI configuration module loaded MST devices:------------/dev/mst/mt4099_pciconf0 - PCI...
View ArticleRDMA Vs IPoIB question
I have a question that whether user applications can choose to adopt RDMA or IPoIB, when IPoIB is enabled? In other words, when IPoIB is enabled, can we still use the RDMA effect while communicating...
View ArticleReceive Side Scaling
Hello,I'm running RHEL 7 acting as a router and performing throughput tests.My machine is a 6 physical cores Xeon with Connect-X 3 EN dual-port.The Spirent generator sends packets from many different...
View ArticleMellonex QSFP cables - what is the difference between the QDR IB cables and...
Hi As per the title?! Is there any reason why I can't use a cable that is tagged as IB for an ethernet QSFP port - obviously with an Ethernet connect-x card? I realise that this will not be...
View ArticleClik here to view.
InfiniBand: Mellanox Technologies MT25208 [InfiniHost III Ex] (rev a0) vs...
Hi Guys I really need some help here, I'm setting up a small cluster and decided instead of using 10gig ethernet I'll play with infiniband SO here I am at step one with 3 * Cisco SFS-HCA-250-A1...
View ArticleRe: RDMA Vs IPoIB question
Hi, Sure - IPoIB operates on a higher layer and you could just tell your application specifically to use the IPoIB interface (ibx) or ignore it and only specify the Mellanox NIC (mlx4 or mlx5) and call...
View ArticleClik here to view.
Re: Port 1 IB Port 2 Ethernet - ESXI 5.1 (5.5)
To my best knowledge this driver still does not support different protocol types on different ports
View ArticleFCA version available with the latest HPCX package
Hello, I've installed the latest HPC-X™ Rev 1.2 package from a tarfile for OFED 1.5.3+ on RHEL6u4 system. The User Manual states that the default FCA is v2.5. I was wondering if I need to install or...
View ArticleClik here to view.
Re: FCA version available with the latest HPCX package
Please, refer to README in the HPC-X package in order to see how to enable FCA v3. The clue is 'hcoll' component.In order to detect OSHMEM version, run oshem_info --version command.
View ArticleRe: RDMA Vs IPoIB question
Thanks Ferber. I got your point. it generally depends that whether my application supports VERBS interface to support RDMA or not. If yes, the it is possible to access the RDMA protocol even if my...
View ArticleHow Do I change the port mode on an MHGH28-XTC
Hi, I have been looking all over to find a command to change the port mode form IB to EN on the MHGH28-XTC cards so we can use them for 10gbe. Can someone help with this? Thanks, Winman1024
View ArticleRe: How Do I change the port mode on an MHGH28-XTC
MHGH28-XTC Looks to be IB DDR hca and is not capable of switching to Ethernet. Regardless, the general command in linux would be "/sbin/connectx_port_config [-d|--device <PCI device id>]...
View ArticleRe: How Do I change the port mode on an MHGH28-XTC
These cards are capable of either DDR IB or 10Gbe. Thank you, Winman1024
View ArticleClik here to view.
How to fix SRP initiator errors with MLNX OFED 2.3-2.0.5 and RHEL6.5 0- how...
I see a lot of scsi errors with srp when I connect from RHEL 6.5 to a DDN S2A 9900 storage system: kernel: scsi host9: SRP abort calledkernel: scsi host9: Null scmnd for RSP w/tag 0000000000000005 The...
View ArticleRe: MLNX_OFED_LINUX-1.5.3-3.0.0 kernel error.
Hi Ferber, We are using cm on IPoIB interfaces. Can we change buffer size in production environment and how to do that?Actually, Our environment has different QDR and FDR network and there are some bad...
View ArticleClik here to view.
Re: Port 1 IB Port 2 Ethernet - ESXI 5.1 (5.5)
Unfortunatly, there is nothing to do that.I think this is derive from ESXi's compact hypervior contstructions.Therefore Mellanox must decide can't support some functions on compact ESXi's hypervior.
View ArticleClik here to view.
Re: MLNX_OFED_LINUX-1.5.3-3.0.0 kernel error.
Hi, It can be done by creating a conf file under /etc/modprobe.d/ and specifying the ring buffer value that you'd like to use. for example : /etc/modprobe.d/ib_ipoib.conf :options ib_ipoib...
View Article