Beruflich Dokumente
Kultur Dokumente
It
provides an overview of the Snapshots and Replication features, examines their architectures, use cases
and capabilities. Snapshot and replication creation of LUNs and file systems are covered and their specific
operations are performed.
Note: Snapshots are not full copies of the original data. It is recommended that you do not rely on
snapshots for mirrors, disaster recovery, or high-availability tools. Because snapshots of storage
resources are partially derived from the real-time data in the relevant storage resource, snapshots can
become inaccessible (not readable) if the primary storage becomes inaccessible.
If the snapshot is writable, (click) any writes are handled in a similar manner; stripe space is allocated
from the parent pool and the writes are redirected in 8 kilobyte chunks to the new space. (click) Reads of
newly written data are also services from the new space.
(click) Storage space is needed in the pool to support snapshots as stripes are allocated for redirected
writes. (click) Because of the on-demand stripe allocation from the pool, snapped thick file systems will
transition to thin file system performance characteristics.
It is also possible to copy a snapshot. (click) In this example, the 4 o’clock snapshot is copied and other
than having a unique name, the copy will be indistinguishable from the source snapshot and both capture
identical data states.
(click) Multiple hosts can be attached to any specific LUN Snapshot or multiple snapshots within the tree.
When a host is attached to a Snapshot for access to its data, the attach can be defined for read-only
access or read/write access. In the example, a host attaches to the 3 o’clock Snapshot for read-only
access and will remain unmodified from it’s original snapped data state. A different host will be attached to
the 4 o’clock Snapshot copy for read/write access. By default, the system will optionally create a copy of
the snapshot to preserve its original data state. (click) When the snap is read/write attached, its data state
is marked as modified from its source.
(click) It is also possible to nest copied read/write attached snapshots that form a hierarchy of snapshots
to a maximum of 10 levels deep.
(click) Snapshots of a file system can be created either as read-only or read/write and are accessed in
different manners which will be covered later.
(click) Copies of snapshots are always created as read/write snapshots. The read/write snapshots can be
shared by creating an NFS export or SMB share to them. When shared, they are marked as modified to
indicate their data state is different from the parent object.
(click) It is also possible to nest copied and shared snapshots that form a hierarchy of snapshots to a
maximum of 10 levels deep.
For Block, the snapshot is created on a LUN or a group of LUNs in the case of a Consistency Group. For
File, the snapshot is configured on a file system. For VMware the storage resource is either going to be a
LUN for a VMFS datastore or a file system for an NFS datastore. When creating each of these storage
resources, the Unity system provides a wizard for their creation. Each wizard provides an option to
automatically create snapshots on the storage resource. Each resource snapshot creation is nearly
identical to the other resources.
For storage resources already created, snapshots can be manually created for them from their Properties
page. As with the wizard, the snapshot creation from the storage resource Properties page is nearly
identical to the other resources. The following few slides will show snapshot creation within the Block
storage LUN creation wizard and the File storage file system creation wizard. It will also show creating
manual snapshots from the LUN and file system Properties pages.
Video demonstrations will be provided showing all forms of storage resource snapshot creation.
(click) The wizard contains a dropdown selection that has three different system defined schedules to
select from to create the LUN snapshots. Within each schedule there is also a snapshot retention value.
(click) A customized schedule can also be created for use. The scheduler provides the ability to configure
a snapshot frequency by the hour, day or week. A snapshot retention policy can also be defined.
Note: fields annotated with a red asterisk are required for the configuration.
(click) Select the + icon to create a snapshot of the LUN. The snapshot must be configured with a name;
by default the system provides a name having a year, month, day, hour, minute, second format.
Customized names can also be configured. A Description field for the snapshot can optionally be
annotated. One of three Retention Policies must be configured. The default retention configuration is the
Pool Automatic Deletion Policy which will automatically delete the snapshot if Pool space reaches a
specified capacity threshold. A customized retention time can alternately be selected and configured for
snapshot deletion on a specified calendar day and time. The other alternative is to select the No Automatic
Deletion option should the snapshot need to be kept for an undetermined amount of time.
(click) The wizard contains a dropdown selection that has three different system defined schedules to
select from to create the file system snapshots. Within each schedule there is also a snapshot retention
value.
(click) A customized schedule can also be created for use. The scheduler provides the ability to configure
a snapshot frequency by the hour, day or week. A snapshot retention policy can also be defined.
As noted before, fields annotated with a red asterisk are required for the configuration.
(click) Select the + icon to create a snapshot of the file system. The snapshot must be configured with a
name; the system by default provides a name based on the creation time in a year, month, day, hour,
minute, second format. Customized names can also be configured. An optional Description field for the
snapshot can be annotated. One of three Retention Policies must be configured. The default retention
configuration is the Pool Automatic Deletion Policy which will automatically delete the snapshot if Pool
space reaches a specified capacity threshold. A customized retention time can alternately be selected and
configured for snapshot deletion on a specified calendar day and time within a year of creation. The other
alternative is to select the No Automatic Deletion option should the snapshot need to be kept for an
undetermined amount of time.
The Access Type section requires configuration by selecting one of the two options for the snapshot to be
created read-only or read/write.
(click) Additional user defined schedules can be created as shown here. This example shows a schedule
that will create weekend snapshots that will be created at midnight and retained for two weeks.
(click) Before performing a restore operation, detach hosts attached to any of the LUN snapshots. (click)
Also ensure that all hosts have completed all read and write operations to the LUN you want to restore.
(click) Finally disconnect any host accessing the LUN. This may require disabling the host connection on
the host-side.
(click) Now the Restore operation can be performed. From the 4 o’clock snapshot select the Restore
operation. (click) The system will automatically create a snapshot of the current 5 o’clock data state of the
LUN to capture its current data state before the restoration operation begins. (click) The LUN is restored
to the 4 o’clock data state of the snapshot.
The hosts can now be reconnected to the resources they were connected to prior to the restore and
resume normal operations.
(click) Before performing an Attach to host operation, the host being attached will need to have
connectivity to the Unity array.
Now the Attach operation can be performed. (click) The first step is to select a snapshot to attach to.
(click) The next step is to select an Access Type, either Read-Only or Read/Write. (click) Then the host
or hosts are selected to be attached. (click) Next, the system will optionally create a copy of the snapshot
if a Read/Write Access Type was selected. Thus preserving the data state of the snapshot prior to attach.
(click) Finally, the selected host is attached to the snapshot with the Access Type selected.
(click) Before performing an Detach operation, allow any outstanding read/write operations of the
snapshot attached host to complete.
(click) Now the Detach operation can be performed. From the 3 o’clock snapshot, select the Detach from
host operation. (click) The secondary host is detached from the 3 o’clock snapshot of the LUN.
(click) Before performing a restore operation, disconnect clients from any of the file system snapshots.
(click) Also quiesce IO to the file system being restored.
(click) Now the Restore operation can be performed. From the 4 o’clock snapshot select the Restore
operation. (click) The system will automatically create a snapshot of the current 5 o’clock data state of the
file system to capture its current data state before the restoration operation begins. (click) The file system
is restored to the 4 o’clock data state of the snapshot.
The connections and IO to the resources can now be resumed for normal operations.
(click) A host has to have connectivity to the storage, either via fibre channel or iSCSI, and be registered.
(click) Next, from the Snapshots tab, a snapshot is selected and the snapshot operation Attach to host
needs to be performed.
Now tasks from the host will need to be done. (click) The host will need to discover the disk device that
the snapshot presents to it. (click) Once discovered, then the host can access the snapshot as a disk
device.
(click) On the storage system an NFS and/or SMB share will have to be configured on the read/write
snapshot of the file system. This task is done from their respective pages.
Now tasks from the client will need to be done. (click) The client will need to be connected to the
NFS/SMB share of the snapshot. (click) Once connected, then the client can access the snapshot shared
resource.
A short video follows that demonstrates client access to a file system snapshot.
(click) The first task for an NFS client is to connect to an NFS share on the file system. (click) Access to
the read-only snapshot is established by accessing the snapshot’s hidden .ckpt data path. This path will
redirect the client to the point-in-time view that the read-only snapshot captures.
(click) Similarly, the first task for an SMB client is to connect to an SMB share on the file system. (click)
Access to the read-only snapshot is established by the SMB client accessing the SMB share’s Previous
Versions tab. This will redirect the client to the point-in-time view that the read-only snapshot captures.
(click) Because the read-only snapshot is exposed to the clients through the CVFS mechanism, the
clients are able to directly recover data from the snapshot without any administrator intervention. For
example, if a user either corrupted or deleted a file by mistake, that user could directly access the read-
only snapshot and get an earlier version of the file from the snapshot and copy it to the file system to
recover from.
A short video follows that demonstrates client access to a file system read-only snapshot.
Remote Replication is one method that enables data centers to avoid disruptions in operations. In a
disaster recovery scenario, if the source site becomes unavailable, the replicated data will still be available
for access from the remote site. Remote Replication uses a Recovery Point Objective (RPO) which is an
amount of data, measured in units of time to perform automatic data synchronization between the source
and remote systems. The RPO for asynchronous replication is configurable. The RPO for synchronous
replication is set to zero. The RPO value represents the acceptable amount of data that may be lost in a
disaster situation. The remote data will be consistent to the configured RPO value. The minimum and
maximum RPO values are 5 minutes and 1440 minutes (24 hours).
Remote Replication is also beneficial for keeping data available during planned downtime scenarios. If a
production site has to be brought down for maintenance or testing the replica data can be made available
for access from the remote site. In a planned downtime situation, the remote data is synchronized to the
source before being made available and there is no data loss.
The main focus of this training is with remote replication since it has more elements to configure, create
and manage.
Fundamental to remote replication is connectivity and communication between the source and destination
systems. A data connection is required to carry the replicated data and it is formed from Replication
Interfaces. They are IP-based connections established on each system. A communication channel is also
required to manage the replication session. The management channel is established on Replication
Connections. It defines the management interfaces and credentials for the source and destination
systems.
Asynchronous Replication architecture utilizes Unified Snapshots. The system creates two snapshots for
the source storage resource and two corresponding snapshots on the destination storage resource. These
system created snapshots cannot be modified. Based on the replication RPO value the source snapshots
are updated in an alternating fashion to capture the source data state differences, known as deltas. The
data delta for the RPO timeframe is replicated to the destination. After the data is replicated the
corresponding destination snapshot is updated. The two corresponding snapshots capture a common data
state, known as a common base. The common base can be used to restart a stopped or interrupted
replication session.
(click) The first step of the initial process for asynchronous replication is to create a storage resource of
the exact same capacity on the destination system. The storage resource is created automatically by the
system and contains no data.
(click) In the next step, corresponding snapshot pairs are created automatically on the source and
destination systems. They capture point-in-time data states of their storage resource.
(click) The first snapshot on the source system is used to perform an initial copy of its point-in-time data
state to the destination storage resource. This initial copy can take a significant amount of time if the
source storage resource contains a large amount of existing data.
(click) Once the initial copy is complete, the first snapshot on the destination system is updated. The data
states captured on the first snapshots are now identical and form a common base.
(click) Because the source storage resource is constantly changing, its data state is no longer consistent
with the first snapshot point-in-time. (click) In the synchronization process, the second snapshot on the
source system is updated, capturing the current data state of the source.
(click) A data difference, or delta is calculated from the two source system snapshots and a differential
copy is made from the second snapshot to the destination storage resource.
(click) After the differential copy is complete, the second snapshot on the destination system is updated
to form a common base with its corresponding source system snapshot.
(click) The cycle of differential copies continues for the session by alternating between the first and
second snapshot pairs based on the RPO value. The first source snapshot is updated, the data delta is
The same fundamental remote replication connectivity and communication between the source and
destination systems seen earlier for asynchronous remote replication are also required for synchronous
replication. A data connection to carry the replicated data is required and is formed using fibre channel
connections between the replicating systems. A communication channel is also required to manage the
replication session. For synchronous replication, part of the management is provided using Replication
Interfaces that are IP based interfaces for SPA and SPB using specific Sync Replication Management
Ports. The management communication between the replicating systems is established on a Replication
Connection. It defines the management interfaces and credentials for the source and destination systems.
Synchronous Replication architecture utilizes Write Intent Logs (WIL) on each of the systems involved in
the replication. These are internal LUNs created automatically by each system. There is a WIL for SPA
and one for SPB on each system. They hold fracture logs that are designed to track changes to the source
LUN should the destination LUN become unreachable. When the destination becomes reachable again it
will automatically recover synchronization to the source using the fracture log, thus avoiding the need for a
full synchronization.
(click) The first step of the initial process for synchronous replication is to create a storage resource of the
exact same capacity on the destination system. The storage resource is created automatically by the
system and contains no data.
(click) In the next step, SPA and SPB Write Intent Logs are automatically created on the source and
destination systems.
(click) An initial synchronization of the source data is then performed. It copies all of the existing data from
the source to the destination. The source resource is available to production during the initial
synchronization but the destination will be unusable until the synchronization completes.
Once the initial synchronization is complete, the process to maintain synchronization begins. (click)
When a primary host writes to the source the system delays the write acknowledgement back to the host.
(click) The write is replicated to the destination system. Once the destination system has verified the
integrity of the data write (click) it sends an acknowledgement back to the source system. (click) At that
point, the source system sends the acknowledgement of the write back to the host. The data state is
synchronized between the source and destination. Should recovery be needed from the destination, its
RPO would be zero.
(click) Should the destination become unreachable, the replication session will be out of synchronization.
The source Write Intent Log for the SP owning the resource will track the changes. (click) When the
destination becomes available the systems will automatically recover synchronization using the WIL.
An Active session state indicates normal operations and the source and destination are In Sync.
A Paused session state indicates the replication has been stopped and will have a Sync State of
Consistent indicating the WIL will be used to perform synchronization of the destination.
A Failed Over session will have one of two Sync States. It can show Inconsistent meaning the Sync State
was not In Sync or Consistent prior to the Failover. If the Sync State was In Sync prior to Failover, it will be
Out of Sync after session Failover.
A Lost Sync Communications session state indicates the destination is unreachable. It can have any of the
following Sync States: Out of Sync, Consistent or Inconsistent.
A Sync State of Syncing indicates a transition from Out of Sync, Consistent or Inconsistent due to the
session changing to an Active state from one of its other states; for example if the system has been
recovered from the Lost Sync Communications state.
For Block, the replication is created on a LUN or a group of LUNs in the case of a Consistency Group. For
File, the replication is configured on a NAS server and file systems. For VMware the storage resource is
either going to be a LUN-based VMFS datastore or a file system-based NFS datastore. When creating
each of these storage resources, the Unity system provides a wizard for their creation. Each wizard
provides an option to automatically create the replication on the resource. Each resource replication
creation is nearly identical to the other resources.
For resources already created, replications can be created manually from their Properties page. As with
the wizard, the replication creation from the resource Properties page is nearly identical to the other
resources. The following few slides will show replication creation within the Block storage LUN creation
wizard and the File storage file system creation wizard. It will also show replications created manually
from the LUN and file system Properties pages.
Video demonstrations will also be provided for the resource replication creation.
For Asynchronous Replication, (click) the Replication Interfaces are dedicated IP-based connections
between the systems (click) that will carry the replicated data. The interfaces are defined on each SP
using IPv4 or IPv6 addressing and will establish the required network connectivity between the
corresponding SPs of the source and destination systems. (click) The Replication Connection pairs
together the Replication Interfaces between the source and destination systems. (click) It also defines the
replication mode between the systems; Asynchronous, Synchronous or both. (click) The connection is
also configured with the management interface and credentials for both of the replicating systems.
(click) The CLI console command can be used to verify the Fibre Channel port that the system has
specified as the Synchronous FC Ports on the SPs. The slide shows an example of running the UEMCLI
command “/remote/sys show –detail” command. In the abbreviated example output, Fibre Channel
Port 4 is specified by the system as the Synchronous FC port for SPA and SPB.
Once the Synchronous FC Ports on the source and destination systems have been verified, Fibre Channel
connectivity can be established between the corresponding ports on the SPs of each system. Direct
connect or zoned fabric connectivity is supported.
Although the Synchronous FC ports will also support host connectivity, it is recommended that they be
dedicated to Synchronous Replication.
From the system’s interface, the Protection and Mobility tree includes an Interfaces option. (click) From
the Interfaces page new Replication Interfaces can be created. On the creation screen, an available
Ethernet Port from the system must be selected. An IP address and subnet mask must be provided for
both SPs. Gateway addressing is optional and a VLAN ID configuration is also provided if needed.
As noted before, Replication Interfaces need to be created on both of the replicating systems. The
creation of Replication Interfaces needs to be repeated on the peer system.
As noted before; there is a dependency between a NAS Server and a file system. The NAS server must
be replicated prior to any associated file system.
In the example NAS Server replication shown, the NAS Server has an associated file system and a
separate replication session will be created for it. The table details the destination resources that will be
used for the file system. (click) It can be selected and edited to further define its destination resources.
In the example, sessions for the (click) NAS Server and its associated file system will be created.
When it is complete the created sessions (click) can be viewed from the Replications page by selecting
the Sessions section.
The Replication Interfaces need to be configured on both the source and destination systems.
In the example, the session (click) settings for the replication and destination are displayed.
When it is complete the created sessions (click) can be viewed from the Replications page by selecting
the Sessions section.
(click) Similar management of a session can also be done from the Properties page of the replicated
object. The Replication tab displays information about the session, provides certain editable fields and
(click) buttons to delete or perform various replication operations.
The example is from an asynchronous file system replication that is in a normal state. From the source it is
possible to perform session Pause, Sync or Failover with Sync operations. From the destination it is only
possible to perform a session Failover operation.
The example illustrates the order of failover for a NAS Server and an associated file system. (click)
Failover must be done first to the NAS Server, (click) then to its associated file system. The same is true
for the Resume operation after Failover. The Resume operation is initiated first on the NAS Server then
the associated file system. Failback is done in the same order; first to the NAS Server then to the
associated file system.
The example illustrates the process of the operation. (click) It starts with issuing the Failover with Sync
operation from Site A which is the primary production site. (click) The operation will remove access to the
replicated object on Site A. (click) A synchronization from the Site A object to the Site B object happens
next and when it completes the replication session is paused. (click) The operation then makes the Site B
object available for access.
The example illustrates the process of the operation. (click) The primary production site becomes
unavailable and all its operations cease. Data is not available and replication between the sites can no
longer proceed. A Failover operation is issued from Site B which is the secondary production site. (click)
The operation Pauses the existing replication session so that the session will not start again should Site A
become available. (click) The operation then makes the Site B object available for access and production
can resume.
The example illustrates the process of a Resume operation for a session that is failed over. (click) The
Site A replicated object must be available before the replication session can be resumed. (click) The
Resume operation is issued from Site B. (click) The operation will restart the Paused session in the
reverse direction. This will update the Site A object with any changes that may have been made to the Site
B object during the failover. This results in the session resuming in the reverse direction and returned to a
normal state. One may use this method of restarting replication rather that a Failback operation if
production had been serviced from the Site B object for a significant amount of time and thus has
accumulated a significant amount of change from the Site A object. Replication in the reverse direction will
synchronize the Site A object to the data state of the Site B object. To return production to the Site A
object would require a session Failover operation followed by another Resume operation.
The example illustrates the process of a Failback operation. (click) The Site A replicated object must be
available before the Failback operation can be initiated on a session. (click) The Failback operation is
issued from Site B. (click) The operation will remove access to the Site B object and synchronize the Site
A object to the data state of the Site B object. (click) The operation then allows access to the Site A object
for production. (click) Replication is restarted using the Site A object as a source and the Site B object as
a destination. This single operation returns the object’s replication state as it was prior to the failover. One
would use this operation to fail back from failovers lasting for only short time periods. It is important to note
that if the Site B object had accumulated a significant amount of change due to long periods of failover, the
resync time can take a significant amount of time.