Dear Sir or Madam,

My name is Meiko Prilop and I am currently writing my master's thesis in the 
field of caching using RDMA at the Vienna University of Technology. In the 
course of my research, I came across your Crail project, which is exactly the 
system I need for my master's thesis. I found your contact in the Github 
project incubator-crail and wanted to try my luck asking questions about the 
project. In the past, I was able to get in touch with Dr Metzler through my job 
at IBM, who was able to give me information regarding Soft-iWARP. However, I am 
now faced with the hurdle of setting up and testing Crail.

For my setup I have Ubuntu instances of version 18.04 running, where I setup 
Soft-iWARP accordingly to the information in

      https://github.com/animeshtrivedi/blog/blob/master/post/2019-06-26-siw.md

I am then prompted three devices using the command ibv_devices:

       device                     node GUID
    ------                    ----------------
    siw_docker0          0242187be7d30000
    siw_lo                   7369775f6c6f0000
    siw_enp1s0             525400afe5890000

And more precisely, using rdma link show:

      1/1: siw_lo/1: state ACTIVE physical_state LINK_UP
               2/1: siw_enp1s0/1: state ACTIVE physical_state LINK_UP
      3/1: siw_docker0/1: state ACTIVE physical_state LINK_UP

I have been able to test rping functionality between two instances so far.

Next, I followed the steps described in:
      https://crail.readthedocs.io/en/latest/docker.html
Although it appears to be the old link to the read me files, the one stated in 
the incubator-crail github are not available anymore. It appears to me that 
these setup descriptions are still the same as in the doc folder tho.

I cloned the repository found at:
      https://github.com/apache/incubator-crail/tree/master
and created an image using the Dockerfile found at /docker/RDMA.

I then try to run the command stated below:

sudo docker run -it --network host -e NAMENODE_HOST=rdma0 -e INTERFACE=enp1s0 
--cap-add=IPC_LOCK --device=/dev/infiniband/uverbs0 
--device=/dev/infiniband/rdma_cm -v /dev/hugepages:/dev/hugepages crail-rdma 
namenode

I further recognized that uverbs0 and rdma_cm are available in the path given.

When running the command, I get this error:

Exception in thread "main" java.io.IOException: j2c::createEventChannel: 
rdma_create_event_channel failed: No such device

        at com.ibm.disni.verbs.impl.NativeDispatcher._createEventChannel(Native 
Method)
        at 
com.ibm.disni.verbs.impl.RdmaCmNat.createEventChannel(RdmaCmNat.java:60)
        at 
com.ibm.disni.verbs.RdmaEventChannel.createEventChannel(RdmaEventChannel.java:67)
        at com.ibm.disni.RdmaCmProcessor.<init>(RdmaCmProcessor.java:48)
        at com.ibm.disni.RdmaEndpointGroup.<init>(RdmaEndpointGroup.java:61)
        at com.ibm.darpc.DaRPCEndpointGroup.<init>(DaRPCEndpointGroup.java:47)
        at com.ibm.darpc.DaRPCServerGroup.<init>(DaRPCServerGroup.java:58)
        at 
com.ibm.darpc.DaRPCServerGroup.createServerGroup(DaRPCServerGroup.java:52)
        at 
org.apache.crail.namenode.rpc.darpc.DaRPCNameNodeServer.init(DaRPCNameNodeServer.java:56)
        at org.apache.crail.namenode.NameNode.main(NameNode.java:92)


Unfortunately, I can't seem to fix this issue at this point. I hope you can 
help me with this problem, Crail seems to me to be a fundamental building block 
for my Master's thesis!

Thank you very much for your time and I look forward to your feedback.


Yours sincerely

Meiko Prilop

Reply via email to