agent hangs using kernel 4.14.36+

[jcvenegas]

Description of problem

When we try to use kata with the latest kata 4.14.x. The container creation hangs at create command.

Expected result

The agent can be created with a new kernel

Actual result

The kata containers master fails to create kata. This does not happend to kata 1.0.0

---

Looking to the cause of the issue grpc request CreateSandbox never finish. This is due to the agent in early versions create a nspath using rand golang module https://github.com/kata-containers/agent/blob/master/namespace.go#L52.

rand is failing not returning due to the low-entropy. This is related to https://github.com/kata-containers/agent/blob/master/namespace.go#L52 as pointed in https://github.com/linuxkit/linuxkit/issues/3032.

For this issue I think we should split in two:

• Remove random dependency at container creation. The ns-paths are only created once so we can change it for a known path.
• Add a test to make sure rand (/dev/random) works in a container.
• Improve rand inside a container (use VirtIORNG or rngd if needed).

level=info msg="[\x1b[0;32m  OK  \x1b[0m] Started Kata Containers Agent.\n" name=kata-proxy pid=27048 source=agent
 level=info msg="[\x1b[0;32m  OK  \x1b[0m] Reached target Kata Containers Agent Target.\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:42.964796872Z\" level=info msg=announce device-handlers=\"blk,scsi\" name=kata-agent pid=101 source=agent storage-handlers=\"ephemeral,9p,blk,scsi\" system-memory=\"2045296 kB\" version=1.0.0-f06eb20527f01097f9185a65a6f77e1ed5d2d360\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:42.986904769Z\" level=info msg=\"Enable grpc tracing\" name=kata-agent pid=101 source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.007601883Z\" level=info msg=\"agent grpc server starts\" name=kata-agent pid=101 source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.012611116Z\" level=debug msg=\"New serial channel event\" events=4 name=kata-agent pid=101 source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.068168302Z\" level=debug msg=\"new request\" name=kata-agent pid=101 req=\"interface:<device:\\\"eth0\\\" name:\\\"eth0\\\" IPAddresses:<address:\\\"172.17.0.2\\\" mask:\\\"16\\\" > mtu:1500 hwAddr:\\\"02:42:ac:11:00:02\\\" > \" request=/grpc.AgentService/UpdateInterface source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.073116342Z\" level=info msg=\"Getting interface from MAC address\" interface-name=eth0 mac-address=\"02:42:ac:11:00:02\" name=kata-agent pid=101 source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.103680142Z\" level=info msg=\"Link found\" interface-name=eth0 link=\"&{LinkAttrs:{Index:2 MTU:1500 TxQLen:1000 Name:eth0 HardwareAddr:02:42:ac:11:00:02 Flags:broadcast|multicast RawFlags:4098 ParentIndex:0 MasterIndex:0 Namespace:<nil> Alias: Statistics:0xc4202045d4 Promisc:0 Xdp:0xc4200371c0 EncapType:ether Protinfo:<nil> OperState:down NetNsID:0 NumTxQueues:0 NumRxQueues:0}}\" mac-address=\"02:42:ac:11:00:02\" name=kata-agent pid=101 source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="[    1.340945] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.107904525Z\" level=debug msg=\"request end\" duration=38.233392ms name=kata-agent pid=101 request=/grpc.AgentService/UpdateInterface resp=\"device:\\\"eth0\\\" name:\\\"eth0\\\" IPAddresses:<address:\\\"172.17.0.2\\\" mask:\\\"16\\\" > mtu:1500 hwAddr:\\\"02:42:ac:11:00:02\\\" \" source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.120917838Z\" level=debug msg=\"new request\" name=kata-agent pid=101 req=\"routes:<Routes:<gateway:\\\"172.17.0.1\\\" device:\\\"eth0\\\" > Routes:<dest:\\\"172.17.0.0/16\\\" device:\\\"eth0\\\" source:\\\"172.17.0.2\\\" scope:253 > > \" request=/grpc.AgentService/UpdateRoutes source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.126118417Z\" level=debug msg=\"request end\" duration=4.258152ms name=kata-agent pid=101 request=/grpc.AgentService/UpdateRoutes resp=\"Routes:<gateway:\\\"172.17.0.1\\\" device:\\\"eth0\\\" > Routes:<dest:\\\"172.17.0.0/16\\\" device:\\\"eth0\\\" source:\\\"172.17.0.2\\\" scope:253 > \" source=agent\n" name=kata-proxy pid=27048 source=agent
 level=info msg="time=\"2018-06-26T15:43:43.131865383Z\" level=debug msg=\"new request\" name=kata-agent pid=101 req=\"hostname:\\\"8d61a7c4633f\\\" storages:<driver:\\\"9p\\\" source:\\\"kataShared\\\" fstype:\\\"9p\\\" options:\\\"trans=virtio,version=9p2000.L\\\" options:\\\"nodev\\\" options:\\\"msize=8192\\\" mount_point:\\\"/run/kata-containers/shared/containers/\\\" > storages:<driver:\\\"ephemeral\\\" source:\\\"shm\\\" fstype:\\\"tmpfs\\\" options:\\\"noexec\\\" options:\\\"nosuid\\\" options:\\\"nodev\\\" options:\\\"mode=1777\\\" options:\\\"size=67108864\\\" mount_point:\\\"/run/kata-containers/sandbox/shm\\\" > \" request=/grpc.AgentService/CreateSandbox source=agent\n" name=kata-proxy pid=27048 source=agent

[sboeuf]

@jcvenegas I agree with 1 and 2, but about 3, I thought this was not working (adding virtio-rng-pci). Did you manage to get it working, or do you mention this as an open question ?

[amshinde]

@jcvenegas Sounds good. I had added the randomness in case more than one namespace needed to be created of the same type. But that is not required as of now.

[jcvenegas]

@sboeuf just as open question @amshinde good , lets handle later if is needed to create another namespace of the same type.