The ego.conf file contains the configuration information for the Symphony cluster. The configuration file is also used to connect to a Symphony cluster from a client machine that is not part of the cluster.
Specifies that share ratio is to always be honored when two or more consumers are competing for resources. Whenever consumers compete for resources, resources are reclaimed and distributed in proportion to share ratio. Workload among all consumers is taken into account and resource distribution is adjusted according to workload.
When set to n or undefined, resources are distributed according to configured share ratio but resource distribution/reclaim is not adjusted according to workload. When two consumers compete for resources, consumers that have not received up to their share ratio can only reclaim resources up to their share ratio. When consumers have reached their share ratio, distribution of additional resources to consumers is done in First-Come, First-Served order.
Limits the maximum size of the allocation event data file (named ego.stream by default) where the event logger stores event data. When a data file exceeds this size, the events logger archives the file and creates a new data file. The events logger maintains one archive file and overwrites the old archive with the new archive.
If your system logs a large number of events, you should increase the maximum file size to see more archived event data. If your disk space is insufficient for storing these files, you should decrease the maximum file size, or change the file path to a location with sufficient storage space.
For a production cluster, the average data file switch time, i.e., the time it takes the data file to reach its maximum size and switch to the archive, should be greater than 3 minutes. If the switch time is too frequent, increase the file size. Note that the file size also impacts the data file writing speed so if the file is too large, writing speed will slow down.
Adds time to all grace periods in seconds. A grace period is the time the system waits before reclaiming resources from a borrowing consumer when a lending consumer requests them back. This time period allows any running work to finish running before the resource is reclaimed.
This parameter adds time to any grace period, cluster wide. All consumers have a grace period. The grace period’s default is 120 seconds.
Enables automatic removal of dynamic hosts from the cluster and specifies the timeout value (minimum 10 minutes). To improve performance in very large clusters, you should disable this feature and remove unwanted hosts from the host cache file manually.
Specifies the length of time the system waits for a dynamic host that is unavailable before the master host removes it from the cluster.
When LIM starts on a host, if the master host already recognizes the host (for example, a static host or a dynamic host that had previously started and joined the cluster), it does not need to send a join request to the master LIM. The master LIM sends acknowledgement signals to all hosts that the master host already recognizes.
If LIM does not receive acknowledgement from the master, it will send the join request and wait for the EGO_DYNAMIC_HOST_WAIT_TIME for the acknowledgement again. Therefore, EGO_DYNAMIC_HOST_WAIT_TIME is the interval, in seconds, that the compute host waits for the master LIM acknowledgement so that it can join the cluster.
Once the acknowledgement signal is received, LIM on the compute host will start all other processes such as PEM, ELIM, etc, on the host. If there is no acknowledgement from the master LIM after 20 tries, the local LIM exits, i.e., the cluster join operation fails.
There are two ways to express EGO_DYNAMIC_HOST_WAIT_TIME. You can specify a value that applies to all attempts to join the cluster or you can specify two intervals for EGO_DYNAMIC_HOST_WAIT_TIME. The first interval applies to the first attempt (of the 20 attempts) and the second interval applies to each of the subsequent attempts.
(Optional) If all of the hosts in your cluster can be reached using short host names, you can configure EGO to use the short host names by specifying the portion of the domain name to remove. If your hosts are in more than one domain or have more than one domain name, you can specify more than one domain suffix to remove, separated by a colon (:).
EGO accepts hostA, hostA.foo.com, and hostA.bar.com as names for host hostA, and uses the name hostA in all output. The leading period ‘.’ is required.
In the above example, EGO accepts hostA, hostA.platform.com, and hostA.generic.com as names for hostA, and uses the name hostA in all output.
Setting this parameter only affects host names displayed through EGO, it does not affect DNS host lookup.