Support Policies for RHEL High Availability Clusters - Management of Highly Available Filesystem Mounts
Contents
Overview
Applicable Environments
- Red Hat Enterprise Linux (RHEL) with the High Availability Add-On
Useful References and Guides
Introduction
This guide offers Red Hat's policies and requirements around managing filesystem mounts in a RHEL High Availability cluster. Users of RHEL High Availability clusters should adhere to these policies in order to be eligible for support from Red Hat with the appropriate product support subscriptions.
Policies
Supported filesystems with Red Hat's ocf:heartbeat:Filesystem or fs resource-agents: Red Hat's ocf:hearbeat:Filesystem resource-agent for pacemaker clusters and fs resource-agent for rgmanager clusters are not limited by any static list of filesystem types that can be used. Certain targeted filesystem types are designed with special handling built-in to the agent where appropriate, but in general the agent is not limited to only a certain set of filesystem types. It may be possible to mount many different filesystem types with these agents and have them managed properly.
However, Red Hat does not provide support for every filesystem type that may be mountable with these agents, and does not test every combination. Red Hat's support for the RHEL High Availability product under High Availability Add-On subscriptions is limited to the operation of the agent and interaction with it by the resource manager (pacemaker or rgmanager). If a concern should arise with the underlying filesystem while it is managed by the cluster and one of these resource types, the assistance that can be provided by Red Hat for that concern is subject to its policies and scope of coverage for that filesystem type. Red Hat may have policies or requirements documented regarding various filesystems, including (but not necessarily limited to) the following:
With filesystems that are supported by Red Hat in RHEL, Red Hat's support team for that component can be engaged to assist, subject to the policies and requirements around that component. With filesystems that Red Hat does not support, Red Hat's direct assistance will be limited to the operation of the resource agent and cluster software, and either the vendor of that filesystem technology or community sources may need to be engaged to assist with the filesystem itself.
These cluster resource-agents are not guaranteed to be able to manage every filesystem type in existence. The agents are likely to be able to handle standard filesystem implementations that can be enabled with the mount command, are listed in /proc/filesystems of the cluster members, and can be interacted with using standard shell commands (cd, ls, touch, dd, etc). If using one of these resource-agents with a filesystem type that Red Hat does not support, it is up the customer organization to test that setup and determine if the agent is capable of managing that mount properly and safely.
If a concern is detected with the interaction between the resource-agent and a certain filesystem type, Red Hat is interested to hear this concern and consider whether that type can be accommodated. Red Hat cannot guarantee the agent will be developed for any particular filesystem type, but Red Hat Support can accept feedback and may be able to offer suggestions or improvements, at its own discretion on a commercially reasonable basis.
Supported filesystems with Red Hat's rgmanager clusterfs resource-agent: Red Hat's clusterfs agent for rgmanager clusters is only designed to manage gfs or gfs2 filesystem types. This agent is not intended and will not handle other types, and Red Hat cannot assist with it in conjunction with any other type. Red Hat's support for clusterfs resources with gfs2 is subject to its other policies covering that component and product, within Red Hat's scope of coverage.
Supported filesystems with Red Hat's rgmanager netfs resource-agent: Red Hat's netfs agent for rgmanager clusters is only designed to manage nfs, nfs4, and cifs filesystem types. This agent is not intended and will not handle other types, and Red Hat cannot assist with it along in conjunction with any other type. Red Hat's support for netfs resources with any of these types is subject to its other policies covering those types, within Red Hat's scope of coverage.
No mounting NFS filesystems exported by this cluster: Due to technical issues in the NFS protocol, Red Hat does not provide support for "loopback mounting" NFS exports - where the same hosts exports an NFS share and mounts it too.
For this reason, Red Hat does not support exporting an NFS share from any cluster member and mounting it via any resource managed by the RHEL High Availability cluster software. There is no distinction whether the export is managed by the resource manager or if its presented directly through an individual host's NFS system - in either situation Red Hat cannot provide assistance with such NFS exports being mounted by a managed cluster resource.
This policy applies to managed Filesystem resources, netfs resources, systemd services or mounts running as resources, lsb scripts running as resources, or any other managed resource type.
This policy applies even if the export and managed resource are configured in such a way that they are expected to run on separate hosts in the cluster. If they come from the same cluster, Red Hat cannot assist with the mount.
A gfs2 filesystem only supports a maximum of 16 nodes: The maximum number of cluster nodes that can mount a gfs2 filesystem is 16 nodes even if the cluster supports more members in the cluster. See the following document for the support limits for gfs2: Global File System 2 | 1.1. GFS2 Support Limits.
Multi Mount Protection: Multi Mount Protection (MMP) is not supported on filesystems that are managed by pacemaker. pacemaker can achieve similar protection from multiple cluster nodes writing to a shared filesystem by letting pacemaker manage a filesystem on HALVM (which allows only 1 cluster node to mount the filesystem at a time).
Comments