mdsd − Pstorage metadata server (MDS) daemon
mdsd [OPTIONS]
pstorage−mdsd start|stop|reload|restart|status [CLUSTER_NAME|DIR]
mdsd(1) is a Pstorage metadata server (MDS) daemon. This daemon is responsible for managing cluster metadata (file names, directory structures, chunk locations, etc.) and should be run in multiple instances for a higher cluster availability.
mdsd(1) can be started using the monitor service pstorage−mdsd. This service accepts different commands for configuring and checking the status of MDSs that belong to either all configured clusters on a host, only the specified cluster or a single MDS.
Normally, all mdsd(1) parameters are specified in the /etc/pstorage/mds.config file used by pstorage−mdsd. The parameters are listed below for your reference only.
−a PROTO
mdsd(1) authentication protocol (e.g., ssl).
−d LEVEL
Set the log level. The configuration parameter mds.log.level is used to configure the current log level for the specified cluster on the fly.
−D
Daemonize the service after start.
−e
Enable the error simulator.
−l LOGFILE
Set the file for storing logs.
−r DIR
Set the mdsd(1) journal directory.
−u USER
Switch to the specified user after start.
−v
Increase the log level.
Metadata information is very critical for providing data access as this information is used by clients when they look for data in the cluster. To make metadata information highly available, MDS supports metadata replication across multiple running MDS instances. A majority of instances is required for the MDS service to be available. The recommended configuration includes 5 MDS instances, which allows you to survive the loss of up to 2 MDSs without clients noticing it.
MDS instances can be added to a cluster using the pstorage make−mds command. To create the first MDS instance, the −I option must be additionally specified.
Non−functioning MDS instances should be removed as soon as possible using the pstorage rm−mds command (e.g., whenever one malfunctioning instance is replaced with a new one) to ensure that all instances are up and running and the majority is not lost on a new instance failure. Here is an example for this: 3 instances are running in the cluster; 1 instance fails; 1 new instance is added. The total number of instances is 4, with one instance offline. Now if one more instance fails, the cluster will become unavailable because the majority (3 running instances) cannot be achieved anymore (only 2 instances out of 4 will be running).
By default, pstorage−mdsd creates two log files for a single mdsd(1) daemon:
• DIR/logs/log
• DIR/logs/fatal−log
The former is filled by mdsd(1) itself, and the latter is filled by the monitor service with fatal messages from mdsd(1).
0
Success
Non−zero
Failure (syntax or usage error; configuration error; cluster failure; unexpected error).
Copyright © 2011−2013, Parallels, Inc. All rights reserved.
pstorage−make−mds(1), pstorage−overview(7), pstorage−config−files(7)