; $Id$ ; Test parameters [test] sleep_time_distr=500-3000 [netcache_client] ; Worker node daemon parameters [server] ; Maximum number of jobs(threads) can be served simultaneously. ; This parameter defines job parallelism. For computationally intensive ; algorithms this value should not be more than number of CPUs ; if set to 'auto', the node will determine the number of CPUs on the ; system and use this number. ; Default value is 'auto' ; max_threads=4 ; initial number of threads created for incoming jobs ; Default number is 1. ; init_threads=1 ; TCP/IP and UDP port number for control messages (like shutdown, version) ; and job notifications. It runs special control thread for incoming ; administrative requests (from netschedule_control and netschedule_admin) ; Can take a ports range (ex. 9300-9310). In this case the system will try ; to find an available port in the given range control_port=9300 ; Server side logging. A worker node can ask its context if this flag is ; set to true log=true ; Internal. ; Delay in seconds node task dispatcher waits for free ; space in the job queue. Lower value usually gives better response ; to shutdown command (CPU traded off) thread_pool_timeout=5 ; Time worker node spends waiting for new jobs without connecting to ; the netschedule server queue. Server sends UPD requests to wake the ; node up. Bigger values of this parameter reduces the netschedule server ; load in expense of job delivery latency (because of ; potential loss of UDP packages) ; job_wait_timeout=10 ; The max total number of jobs after which the node will shutdown itself. ; Restarting the node periodically is useful due to accumulating heap ; fragmentation possible leaks etc. ; default is 0 - means unlimited number of jobs. ; max_total_jobs = 100 ; The max number of failed jobs after which the node will shutdown itself. ; default is 0 - means unlimited number of failed jobs. ; max_failed_jobs = 3 ; When true, server transforms into a daemon, detached from the ; current program group (for UNIX only) daemon=true ; The list of worker nodes which this node will check before attempting ; to retrieve a job from the NetSchedule queue. If at least one of these ; worker nodes has an ideal thread, this node will not connect to the queue ; for a job. This node and all nodes from the given list must be connected ; to the same NetSchedule service, the same queue and must run the same job. ; If the list is empty (defult) then this node is a master. ;master_nodes = service3:9300, service2:9300 ; List of network hosts which are allowed admin access to the worker node ; if this worker node is controled by grid_node_watcher.sh don't forget to ; to add "localhost" to this list. By default any host is allowed. ;admin_hosts = localhost service1 service2 service3 ; Time delay (in seconds) between the node enters an idle mode (all jobs are done ; and there are no new jobs in the queue) and the idle task gets executed. ; Can not be less then 1 sec. Default value is 30 sec. ;idle_run_delay = 30 ; The node will automatically shut itself down if it is idle for ; a continuous period of time longer than this (in seconds): ; Default: 0 - means never auto shutdown ; auto_shutdown_if_idle = 20 ; Specifies if the framework should reuse an instance of the job class. ; Setting this parameter to true means that only one instance of the job class ; will be create per each execution thread. False means that an instance of job ; class will be created per each incoming job. ; Default value is false. ;reuse_job_object = true ; Allows the node to detect infinite loops in the job executions. If a job is being ; executed more then the specified time it is assumed to be in the infinite loop. ; It this happened the node enters in the shutdown mode. Then all other jobs, ; which are being running on this node, are done or also are in infinite loops ; the node will kill itself. ; Default value is 0. This means that the node will not detect inifinit loops. ;infinite_loop_time=30 ; Time in seconds. Specifies how often the node should check states of jobs ; it is processing. It is used as a feedback from the client to see if it wants ; to cancel the job execution check_status_period = 4 [gw_debug] ; Prefix for all debug files run_name = debug_run ; Whether to gather requests or to execute them ;mode = gather ; Whether to gather requests or to execute them ; default is 1. ; Input blobs will be dumped to "..[1..N].inp" ; Output blobs will be dumped to "..[1..N].out" ; Messages will be dumped to "..[1..N].msg" ; Diagnostics will be written to "..log" ;gather_nrequests = 2 ; Whether to gather requests or to execute them ;mode = execute ; Process requests from the following files ; [no default] ; Input blobs will be read from "" ; Output blobs will be dumped to "__.execute.out" ; Messages will be dumped to "__.execute.msg" ; Diagnostics will be written to "__.execute.log" ;execute_requests = [ ...] ; NetSchedule client configuration ; [netschedule_api] ; How to introduce this application to NetSchedule. client_name=test_node ; Specifies NetSchedule service. Format: lb_service_name or host:port service=NS_Test ; Name of the queue (DO NOT use default queue for your application) queue_name=test ; Instructs the worker noded to try to fit its output onto the netschedule's ; internal storage (its size is about 256Kb). If the data does not fit there ; it will be saved into an external blob storage (netcahe). ; Default value is false ;use_embedded_storage = true ; Time interval in seconds when client is automatically rebalanced rebalance_time=10 ; Number of requests before rebalancing rebalance_requests=100 ; Sets a communication timeout (in seconds) between a client and a server ; Default is 12 secs ;communication_timeout = 20 ; Network storage (NetCache) configuration ; [netcache_api] ; How to introduce this application to NetCache. client_name=test_node ; LB service name service=NC_Test ; If this parameter is set to true then before the job is processed the input data blob ; will be stored into a file on the local file system and the job will get a stream to ; that file. ; Default is false cache_input = false ; If this parameter is set to true then the job will get a stream to file on the local ; file system as an output stream. After job is done the content of this file will be ; stored to a blob in the storage. ; Default is false cache_output = false ; If 'cache_input' or 'cache_output' are set to true, sets a temporary directory where ; cached input/ouput is stored. ; Default: . ; tmp_dir=/tmp ; Synonym for tmp_dir. if both are set tmp_dir is used. ; tmp_path=/tmp ; Time interval in seconds when client is automatically rebalanced rebalance_time=10 ; Number of requests before rebalancing rebalance_requests=100 ; Sets a communication timeout (in seconds) between a client and a server ; Default is 12 secs ;communication_timeout = 20