After installing PE, you can change some default settings to further configure the orchestrator and pe-orchestration-services.
Configure the orchestrator and pe-orchestration-services
There are several optional parameters you can add to configure the behavior of the orchestrator and pe-orchestration-services. Because they are profile classes, you can change these in the console in the PE Orchestrator group.
- Integer representing the number of simultaneous task or plan actions that can run at the same time in the orchestrator. All task and plan actions are limited by this concurrency limit regardless of transport type (WinRM, SSH, PCP).
- An integer that determines the maximum number of simultaneous task or plan requests orchestrator can make to bolt-server. Only task or plan executions on nodes with SSH or WinRM transport methods are limited by this setting because only they require requests to bolt-sever.
CAUTION: Do not set a concurrency limit that is higher than the bolt-server limit. This can cause timeouts that lead to failed task runs.
- Disable or enable the PXP service by setting
false. If you disable this setting you can’t use the orchestrator or the Run Puppet button in the console.
- An integer that determines how many concurrent compile requests can be outstanding to the primary server, across all orchestrator jobs.
- Integer that represents the number of days before job reports are removed.
- An integer that represents how many seconds must pass while an agent attempts to connect to a PCP broker. If the agent can’t connect to the broker in that time frame, the run times out.
- Disable or enable orchestration services. Set
- An integer that defines how many times an
orchestrator job allows status requests to time
out before a job is considered failed. Status
requests wait 12 seconds between timeouts, so
multiply the value of the
allowed_pcp_status_requestsby 12 to determine how many seconds the orchestrator waits on targets that aren’t responding to status requests.
Specifies the heap size, or, the amount of memory that each Java process is allowed to request from the operating for the orchestrator to use.
- An integer that determines the maximum number
of JRuby instances that the orchestrator creates
to execute plans. Because each plan uses one JRuby
to run, this value is effectively the maximum
number of concurrent plans. Setting the
orchestrator heap size (
java_args) automatically sets the
jruby_max_active_instancesusing the formula
java_args/ 1024. If the value equals less than one, the default is one JRuby instance.
Configure the PXP agent
Puppet Execution Protocol (PXP) is a messaging system used to request tasks and communicate task statuses. The PXP agent runs the PXP service and you can configure it using Hiera or the console.
- Controls how frequently (in seconds) PXP agents will ping PCP brokers. If the brokers don’t respond, the agents try to reconnect.
- A string that represents the path to the PXP agent log file and can be used to debug issues with orchestrator.
- The amount of time to keep records of old Puppet or task runs on agents. You can declare time in minutes (30m), hours (2h), and days (14d).
- Controls how long tasks are cached after use. You can declare time in minutes (30m), hours (2h), and days (14d).
- Sets a proxy URI used to connect to the pcp-broker to listen for task and Puppet runs.
- Sets a proxy URI used to connect to the primary server to download task implementations.
- Sets the message size, in mb, for pcp_broker, pxp_agent, and the orchestrator. The maximum message size cannot be higher than the default of 64mb, so you can only reduce it.
64(mb)Note: We do not recommend changing the
pcp_max_message_size_mbparameter if you send or receive large payloads because it might cause errors for large task and plan run parameters and output.
Correct ARP table overflow
In larger deployments that use the PCP broker, you might encounter ARP table overflows and need to adjust some system settings.
Overflows occur when the ARP table—a local cache of IP address to MAC address resolutions—fills and starts evicting old entries. When frequently used entries are evicted, network traffic will increase to restore them, increasing network latency and CPU load on the broker.
A typical log message looks like:
[root@s1 peadmin]# tail -f /var/log/messages Aug 10 22:42:36 s1 kernel: Neighbour table overflow. Aug 10 22:42:36 s1 kernel: Neighbour table overflow. Aug 10 22:42:36 s1 kernel: Neighbour table overflow.
To work around this issue:
# Set max table size net.ipv6.neigh.default.gc_thresh3=4096 net.ipv4.neigh.default.gc_thresh3=4096 # Start aggressively clearing the table at this threshold net.ipv6.neigh.default.gc_thresh2=2048 net.ipv4.neigh.default.gc_thresh2=2048 # Don't clear any entries until this threshold net.ipv6.neigh.default.gc_thresh1=1024 net.ipv4.neigh.default.gc_thresh1=1024