I have set up access to HDFS using httpfs setup in Kubernetes as I need to have access to HDFS data nodes and not only the metadata on the name node. I can connect to the HDFS using Node port service with telnet, however, when I try to get some information from HDFS - reading files, checking if the files exist, I get an error:
[info] java.net.SocketTimeoutException: Read timed out
[info] at java.net.SocketInputStream.socketRead0(Native Method)
[info] at java.net.SocketInputStream.socketRead(SocketInputStream.java:116)
[info] at java.net.SocketInputStream.read(SocketInputStream.java:171)
[info] at java.net.SocketInputStream.read(SocketInputStream.java:141)
[info] at java.io.BufferedInputStream.fill(BufferedInputStream.java:246)
[info] at java.io.BufferedInputStream.read1(BufferedInputStream.java:286)
[info] at java.io.BufferedInputStream.read(BufferedInputStream.java:345)
[info] at sun.net.www.http.HttpClient.parseHTTPHeader(HttpClient.java:735)
[info] at sun.net.www.http.HttpClient.parseHTTP(HttpClient.java:678)
[info] at sun.net.www.protocol.http.HttpURLConnection.getInputStream0(HttpURLConnection.java:1587)
What could be the reason for this error? Here is the source code for setting up the connection to HDFS file system and checking files existence:
val url = "webhdfs://192.168.99.100:31400"
val fs = FileSystem.get(new java.net.URI(url), new org.apache.hadoop.conf.Configuration())
val check = fs.exists(new Path(dirPath))
The directory by the dirPath argument exists on HDFS.
HDFS Kubernetes setup looks like this:
apiVersion: v1
kind: Service
metadata:
name: namenode
spec:
type: NodePort
ports:
- name: client
port: 8020
- name: hdfs
port: 50070
nodePort: 30070
- name: httpfs
port: 14000
nodePort: 31400
selector:
hdfs: namenode
---
apiVersion: v1
kind: ReplicationController
metadata:
name: namenode
spec:
replicas: 1
template:
metadata:
labels:
hdfs: namenode
spec:
containers:
- env:
- name: CLUSTER_NAME
value: test
image: bde2020/hadoop-namenode:2.0.0-hadoop2.7.4-java8
name: namenode
args:
- "/run.sh &"
- "/opt/hadoop-2.7.4/sbin/httpfs.sh start"
envFrom:
- configMapRef:
name: hive-env
ports:
- containerPort: 50070
- containerPort: 8020
- containerPort: 14000
volumeMounts:
- mountPath: /hadoop/dfs/name
name: namenode
volumes:
- name: namenode
emptyDir: {}
---
apiVersion: v1
kind: Service
metadata:
name: datanode
spec:
ports:
- name: hdfs
port: 50075
targetPort: 50075
selector:
hdfs: datanode
---
apiVersion: v1
kind: ReplicationController
metadata:
name: datanode
spec:
replicas: 1
template:
metadata:
labels:
hdfs: datanode
spec:
containers:
- env:
- name: SERVICE_PRECONDITION
value: namenode:50070
image: bde2020/hadoop-datanode:2.0.0-hadoop2.7.4-java8
envFrom:
- configMapRef:
name: hive-env
name: datanode
ports:
- containerPort: 50075
volumeMounts:
- mountPath: /hadoop/dfs/data
name: datanode
volumes:
- name: datanode
emptyDir: {}
UPD: Ping returns such results (192.168.99.100 - minikube ip, 31400 - service node port):
ping 192.168.99.100 -M do -s 28
PING 192.168.99.100 (192.168.99.100) 28(56) bytes of data.
36 bytes from 192.168.99.100: icmp_seq=1 ttl=64 time=0.845 ms
36 bytes from 192.168.99.100: icmp_seq=2 ttl=64 time=0.612 ms
36 bytes from 192.168.99.100: icmp_seq=3 ttl=64 time=0.347 ms
36 bytes from 192.168.99.100: icmp_seq=4 ttl=64 time=0.287 ms
36 bytes from 192.168.99.100: icmp_seq=5 ttl=64 time=0.547 ms
36 bytes from 192.168.99.100: icmp_seq=6 ttl=64 time=0.357 ms
36 bytes from 192.168.99.100: icmp_seq=7 ttl=64 time=0.544 ms
36 bytes from 192.168.99.100: icmp_seq=8 ttl=64 time=0.702 ms
36 bytes from 192.168.99.100: icmp_seq=9 ttl=64 time=0.307 ms
36 bytes from 192.168.99.100: icmp_seq=10 ttl=64 time=0.346 ms
36 bytes from 192.168.99.100: icmp_seq=11 ttl=64 time=0.294 ms
36 bytes from 192.168.99.100: icmp_seq=12 ttl=64 time=0.319 ms
36 bytes from 192.168.99.100: icmp_seq=13 ttl=64 time=0.521 ms
^C
--- 192.168.99.100 ping statistics ---
13 packets transmitted, 13 received, 0% packet loss, time 12270ms
rtt min/avg/max/mdev = 0.287/0.463/0.845/0.173 ms
And for the host and port:
ping 192.168.99.100 31400 -M do -s 28
PING 31400 (0.0.122.168) 28(96) bytes of data.
^C
--- 31400 ping statistics ---
27 packets transmitted, 0 received, 100% packet loss, time 26603ms
My colleague found out that the problem was with docker in minikube. Running this before setting up HDFS on Kubernetes solved the problem:
minikube ssh echo "sudo ip link set docker0 promisc on"
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With