[
https://issues.apache.org/jira/browse/STORM-4116?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
alexis updated STORM-4116:
--------------------------
Description:
I have a test after year 2038 (ex: 2040) to validate my topology is not
affected by Y2038 bug.
+Context:+
* I have installed storm 2.4.0 on each node of my platform
* I updated my platform to date 24/04/2040
* In the storm nimbus configuration I set the following pacemaker
configuration:
## ######################################
### Pacemaker configuration ###
######################################
# Cluster state management. PaceMakerStateStorageFactory to use pacemaker
instead of Zookeeper.
storm.cluster.state.store:
"org.apache.storm.cluster.PaceMakerStateStorageFactory"
# Pacemaker servers and port configuration
pacemaker.servers: ["<alias of storm nimbus>"]
pacemaker.port: 6699
# Minimal number of thread used to monitor topologies lifecycle
pacemaker.base.threads: 10
# Maximal number of thread used to monitor topologies lifecycle
pacemaker.max.threads: 50
# Number of maximum thread for each connected client
pacemaker.client.max.threads: 2
# Thread client timeout
pacemaker.thread.timeout: 10
# Childopts for server
pacemaker.childopts: "-Xmx4096m"
# Authentification if needed (Kerberos, etc ...)
pacemaker.auth.method: "NONE"
pacemaker.kerberos.users: []
Size maximum of message sent by supervisor to pacemaker
pacemaker.thrift.message.size.max: 10485760
* In the storm supervisor, I put the following one:
## ######################################
### Pacemaker configuration ###
######################################
# Cluster state management. PaceMakerStateStorageFactory to use pacemaker
instead of Zookeeper.
storm.cluster.state.store:
"org.apache.storm.cluster.PaceMakerStateStorageFactory"
Pacemaker servers and port configuration
pacemaker.servers: ["<alias of storm nimbus>"]
pacemaker.port: 6699
* I submitted my topology
* The topology is well submitted on the supervisor node
+Observations:+
I checked the <alias of storm-nimbus> from supervisor node and I am able to
ping.
In the nimbus log, I observed that after a certain time the topology is
reassigned because no heartbeat has been received inside Nimbus server from
workers. I checked logs and content of sources and I observed that timestamps
of heartbeats (time_secs and uptime_secs variables) are set into integer.
This issue also been observed on 2.7.0 version.
was:
I have a test after year 2038 (ex: 2040) to validate my topology is not
affected by Y2038 bug.
+Context:+
* I have installed storm 2.4.0 on each node of my platform
* I updated my platform to date 24/04/2040
* In the storm nimbus configuration I set the following pacemaker
configuration:
## ######################################
### Pacemaker configuration ###
######################################
# Cluster state management. PaceMakerStateStorageFactory to use pacemaker
instead of Zookeeper.
storm.cluster.state.store:
"org.apache.storm.cluster.PaceMakerStateStorageFactory"
# Pacemaker servers and port configuration
pacemaker.servers: ["<alias of storm nimbus>"]
pacemaker.port: 6699
# Minimal number of thread used to monitor topologies lifecycle
pacemaker.base.threads: 10
# Maximal number of thread used to monitor topologies lifecycle
pacemaker.max.threads: 50
# Number of maximum thread for each connected client
pacemaker.client.max.threads: 2
# Thread client timeout
pacemaker.thread.timeout: 10
# Childopts for server
pacemaker.childopts: "-Xmx4096m"
# Authentification if needed (Kerberos, etc ...)
pacemaker.auth.method: "NONE"
pacemaker.kerberos.users: []
Size maximum of message sent by supervisor to pacemaker
pacemaker.thrift.message.size.max: 10485760
* In the storm supervisor, I put the following one:
## ######################################
### Pacemaker configuration ###
######################################
# Cluster state management. PaceMakerStateStorageFactory to use pacemaker
instead of Zookeeper.
storm.cluster.state.store:
"org.apache.storm.cluster.PaceMakerStateStorageFactory"
Pacemaker servers and port configuration
pacemaker.servers: ["<alias of storm nimbus>"]
pacemaker.port: 6699
* I submitted my topology
* The topology is well submitted on the supervisor node
+Observations:+
I checked the <alias of storm-nimbus> from supervisor node and I am able to
ping.
In the nimbus log, I observed that after a certain time the topology is
reassigned. I checked logs and content of sources and I observed that
timestamps of heartbeats (time_secs and uptime_secs variables) are set into
integer.
This issue also been observed on 2.7.0 version.
> Heartbeats mechanism is affected by Y2038 bug
> ---------------------------------------------
>
> Key: STORM-4116
> URL: https://issues.apache.org/jira/browse/STORM-4116
> Project: Apache Storm
> Issue Type: Bug
> Affects Versions: 2.4.0
> Reporter: alexis
> Priority: Major
>
> I have a test after year 2038 (ex: 2040) to validate my topology is not
> affected by Y2038 bug.
> +Context:+
> * I have installed storm 2.4.0 on each node of my platform
> * I updated my platform to date 24/04/2040
> * In the storm nimbus configuration I set the following pacemaker
> configuration:
> ## ######################################
> ### Pacemaker configuration ###
> ######################################
> # Cluster state management. PaceMakerStateStorageFactory to use pacemaker
> instead of Zookeeper.
> storm.cluster.state.store:
> "org.apache.storm.cluster.PaceMakerStateStorageFactory"
> # Pacemaker servers and port configuration
> pacemaker.servers: ["<alias of storm nimbus>"]
> pacemaker.port: 6699
> # Minimal number of thread used to monitor topologies lifecycle
> pacemaker.base.threads: 10
> # Maximal number of thread used to monitor topologies lifecycle
> pacemaker.max.threads: 50
> # Number of maximum thread for each connected client
> pacemaker.client.max.threads: 2
> # Thread client timeout
> pacemaker.thread.timeout: 10
> # Childopts for server
> pacemaker.childopts: "-Xmx4096m"
> # Authentification if needed (Kerberos, etc ...)
> pacemaker.auth.method: "NONE"
> pacemaker.kerberos.users: []
> Size maximum of message sent by supervisor to pacemaker
> pacemaker.thrift.message.size.max: 10485760
> * In the storm supervisor, I put the following one:
> ## ######################################
> ### Pacemaker configuration ###
> ######################################
> # Cluster state management. PaceMakerStateStorageFactory to use pacemaker
> instead of Zookeeper.
> storm.cluster.state.store:
> "org.apache.storm.cluster.PaceMakerStateStorageFactory"
> Pacemaker servers and port configuration
> pacemaker.servers: ["<alias of storm nimbus>"]
> pacemaker.port: 6699
> * I submitted my topology
> * The topology is well submitted on the supervisor node
> +Observations:+
> I checked the <alias of storm-nimbus> from supervisor node and I am able to
> ping.
> In the nimbus log, I observed that after a certain time the topology is
> reassigned because no heartbeat has been received inside Nimbus server from
> workers. I checked logs and content of sources and I observed that timestamps
> of heartbeats (time_secs and uptime_secs variables) are set into integer.
> This issue also been observed on 2.7.0 version.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
