Академический Документы
Профессиональный Документы
Культура Документы
application environment hosted on a Dell EMC VMAX storage system using Dell EMC
W HITE PAPER
The information in this publication is provided as is. EMC Corporation makes no representations or warranties of any kind with respect
to the information in this publication, and specifically disclaims implied warranties of merchantability or fitness for a particular purpose.
Use, copy ing, and distribution of any EMC software described in this publication requires an applicable software license.
EMC 2, EMC, the EMC logo, Dell EMC AppSync, and EMC VMAX are registered trademarks or trademarks of EMC Corporation in the
United States and other countries. All other trademarks used herein are the property of their respective owners. Copyright 2016 EMC
Corporation. All rights reserved. Published in the USA., 09/16 White Paper, H13621.1
EMC believes the information in this document is accurate as of its publication date. The information is subject to change without notice.
EMC is now part of the Dell group of companies.
TABLE OF CONTENTS
EXECUTIVE SUMMARY................................................................................................. 5
AUDIENCE ........................................................................................................................................................ 5
INTRODUCTION............................................................................................................ 5
USING THE GUIDELINES .............................................................................................. 5
VMAX CONFIGURATION REQUIREMENTS .................................................................... 6
VMAX Block Storage Considerations............................................................................................................ 6
Ad vanced VMAX3 Features Supported........................................................................................................ 6
VMAX SMI-S Pro vider ..................................................................................................................................... 7
REPURPOSING WORKFLOWS.................................................................................... 23
RESTORE FROM APPLICATION COPIES .................................................................... 26
LUN Level Restores from VP Snap and TF Clone copies ....................................................................... 26
VM Le vel Restore from VP Snap and TF Clone copies ........................................................................... 26
VMAX3 Restores with SnapVX .................................................................................................................... 26
EXECUTIVE SUMMARY
This document prov ides technical information about integrating Dell EMC AppSync software with EMC VMAX storage sy stems,
including the environmental caveats that should be taken into consideration when using both AppSync nc and VMAX. This document also
prov ides guidance while deploying the AppSync nc software in environments where application data resides on a VMAX storage system.
AUDIENCE
This white paper is intended f or Dell EMC customers that are application and/or storage administrators who are currently administering
AppSync nc in their environment and Dell EMC internal f field personnel and partners who assist customers with deploying AppSync.
INTRODUCTION
Dell EMC AppSync is software that enables Integrated Copy Data Management (iCDM) with Dell EMC's primary storage systems.
AppSync simplifies and automates the process of generating and consuming copies of production data. By abstracting the underlying
storage and replication technologies, and through deep application integration, AppSync empowers application owners to satisfy copy
demand f or operational recovery and data repurposing on their own. In turn, storage administrators need only be concerned with initial
setup and policy management, resulting in an agile, frictionless environment. AppSync automatically discovers application databases,
learns the database structure, and maps it through the virtualization layer to the underlying storage LUN. It then orchestrates all the
activities required from copy creation and validation through mounting at the target host and launching or recovering the application.
Supported workflows also include refresh, expire, and restore production.
Replication Technology
TimeFinder Clone
TimeFinder Mirror
TimeFinder Snap
TimeFinder VP Snap
SnapVX
RecoverPoint Local*
RecoverPoint Remote*
RecoverPoint
Local and Remote*
SRDF/S
SRDF/A (non-MS Apps only)
VMAX Support
RAID LUN
Pool LUN
Yes
No
No
Not applicable
Not applicable
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
*Note: Applies to VMAX2 only RecoverPoint with VMAX3 is not currently supported.
Supported
Applications
Oracle on Linux and AIX
Microsoft Exchange
Microsoft SQL
Microsoft file systems
Linux and AIX file systems
VMware Datastores
VMAX BLOCK STORAGE CONSIDERATIONS
As noted in the previous section, AppSync supports TimeFinder and SnapsVX technologies. VMAX2 array s support thin pool based
LUNS f or snap operations, and due to this restriction, AppSync only supports TimeFinder VP Snap technology. Traditional TimeFinder/
Snap which utilizes VDEVs, are not supported with AppSync. TimeFinder/Clone and RecoverPoint based copies, with a VMAX2 backend, are the only technologies which support RAID LUNs. On the VMAX3, SnapVX Clone and Snap are both supported, and both
utilize the default SRP (storage resource pool).
A FA
AppSync 2.2.2.2 and greater offers support for VMAX All Flash Array s; the first true fully all flash based VMAX array . Please reference
The VMAX All Flash Storage Family white paper f or more details here - https://www.emc.com/collateral/white-papers/h14920-introto-v max-af-storage.pdf
Queries all required information, such as mapping, device, storage groups, pools, etc.
Performs active management tasks, such as snapshot and clone sessions creating, refreshing, terminating, restoring, LUN
masking etc.
The SMI-S provider software can be installed on a Windows or Linux, virtual or physical, machine that has physical connectivity to the
VMAX(s) being managed. The SMI-S provider must be registered as a resource within AppSync. There should be at least six
gatekeepers exported from the VMAX being managed to the SMI-S provider host. They should be spread across two fabrics/HBAs for
redundancy .
Up to five arrays using a single SMI-S instance can be supported, however, due to performance considerations, ascertained by multiple f
actors, caution should be used and so the recommendation is one SMI-S provider per local and its remote array. Consider other
applications using the provider, total volumes and copies under management, etc, when determining how many arrays one SMI-S
provider can support.
If the SMI-S Provider and AppSync server share the same host, then the SMI-S provider should be installed first. If AppSync is already
installed, install SMI-S provider and restart the AppSync server services (Dell EMC AppSync Datastore service, Dell EMC AppSync
Security Server, and Dell EMC AppSync Server Service).
MODIFYING CREDENTIALS
If the SMI-S provider credentials are modified by the SMI-S admin, they must also be change within AppSync. Navigate to the Storage
Infrastructure menu, highlight the VMAX, highlight the SMI-S host, and then clicking Edit, in order to change the credentials.
PROVIDER PREFERENCE
AppSync uses the SMI-S provider that was used to initially discover the VMAX array , as the preferred provider. The pref erred order, in
the case of multiple SMI-S providers, can be changed within the AppSync UI by highlighting the provider and clicking Set Preferred.
Either stop the ECOM service on the Windows host, or execute the following command to stop the EMC SMI-S provider service:
2.
3.
Execute the command SYMCLI\bin\symcfg list to list the VMAX array ID(s) to disable
4.
In a plain text file, list the IDs of the array(s) that should not be auto-discovered
5.
Save the file as symavoid in the SYMAPI/config directory (ensure no extension exists, such as the Windows .txt extension, which is hidden by
default)
6.
Rename the SYMAPI/db/symapi_db.bin file to something like symapi_db.bin.old{todays date}. The database will be recreated the next time the
SMI-S provider tries to discover arrays.
7.
Start ECOM service or execute the following command to start the EMC SMI-S provider service:
8.
9.
Perform a discovery for the changes to take effect. Do this by running the discover command in ECIM\ECOM\bin \TestSmisProvider.exe binary
shipped with the SMI-S Provider, or running the Solutions Enabler command SYMCLI\bin\symcfg discover and then running the
SYMCLI\bin\symcfg list command to validate.
STORAGE CONFIGURATION
AppSync requires copy devices to create successful application copies using TimeFinder Clone and VP Snap technologies, or in the
case of VMAX3, SnapVX technology using targetless snapshots. AppSync provides two choices to the storage administrator for
configuring a copy device to be used by AppSync; creating and pairing targets for TimeFinder Clone and VP Snap, and one optional
choice f or SnapVX copies, which are:
Storage Groups: Using a dedicated storage group (one or more), which contain preconfigured target devices
Prov ides thresholds, so if a pool reaches a certain used percentage (90% by default), AppSync will no longer use that
pool to create new devices
VMAX3 array s use the default Storage Resource Pool (SRP) and cannot be configured differently
There are fundamental differences between VMAX2 and VMAX3 array s, when it comes to how storage is configured. With the VMAX2
array s, optionally choose to use storage groups, and/or storage pools. With the VMAX3 array , optionally choose a storage group. There
is also a difference with how VMAX3 uses storage with repurposing workflows versus service plan workflows. Please refer to the
Repurposing Workf low section for more details.
VMAX2
Storage administrators can use either or both options. If neither option is used, AppSync jobs will f ail, because it has no devices in a
storage group to use as target devices, nor does it have the authority to use a storage pool to create new devices. If just the storage group
is defined, then AppSync will attempt to use devices within that storage group, and if none are available, then the process will f ail. When
both options are used, AppSync first tries to get copy storage from previously-provisioned devices (in a storage group defined), and if
none are found, storage pools are used to provision new targets. The devices AppSync provisions are not added to the defined storage
group, so some administrators may wish to manually add them for proper housekeeping, otherwise, the devices AppSync creates are
easily identifiable, as seen in the section titled Device Management on the Backend VMAX and depicted in Figure 12 and Figure 13.
VMAX3
Storage administrators can use an optional storage group f or copy devices. If a storage group is not defined, AppSync will use the SRP to
create targetless/unlinked devices, during service plan workflows, and will create target/linked devices from the SRP during
repurposing workflows. If a storage group is defined, AppSync still uses the SRP during the copy phase of a service plan workflow,
however, will use the devices in the storage group during the mount operation; when a device requires linking. If no device is available in
a storage group defined within AppSync, the mount phase will fail. When running a repurposing workflow with a storage group
st
defined, AppSync requires the target device to be present during the 1 generation copy phase.
Configuring a storage group prevents AppSync from creating devices in the SRP devices must be available in the configured
storage group
10
Serv ice Plans do not require target dev ices during the copy phase, but requires during the mounting phase
Repurposing workf lows require target dev ices for the copy phase for the 1 generation copies, but not for the 2 generation copies
please ref er to the SnapVX Linking section for more details.
st
nd
VMAX3 Behavior
Make sure the target copy device is of the same size and conf iguration (i.e. meta configuration) as source device
For VP Snaps, make sure the target dev ices are thin provisioned
If the source dev ices are already in a VP Snap session f rom a thin pool, make sure the target copy devices are also provisioned in
the same pool (VP Snaps of same source must all be f rom same pool) please ref er to the TimeFinder VP Snap documentation
f or more details
If an application already has an existing TimeFinder clone or VP Snap session, ensure the target devices of that existing session are
placed in one of the storage groups conf igured within AppSync. AppSync reuses existing sessions.
All dev ices in the configured storage group(s) are used as copy devices, so never add production volumes into AppSync assigned
storage group(s) as it will result in data loss. Do not share the same storage group(s) across multiple AppSync servers.
As seen in Figure 5Figure 5 - Selecting Storage Groups and Pools, AppSy nc discovers storage groups which are not part of any
masking v iew. To conf igure a storage group f or an AppSync target device selection, enable it by clicking on the checkbox. AppSync
11
allows selecting multiple storage groups. Once selected, a warning states that the devices within in the selected storage group will be
used as target f or copies created by AppSync, as seen in Figure 6. The data on those dev ices will be ov erwritten.
Rev iew the storage details within AppSy nc, under Storage Infrastructure, by clicking on the VMAX, as seen in Figure 7. Notice that
the VMAX2 shows the selected storage pool, but the VMAX3 does not. The only supported pool on VMAX3 is the def ault SRP. When
multiple storage groups are added, the Number of Devices is the sum of all storage groups.
VMAX2
VMAX3
Conf iguring/de-configuring storage groups is accomplished by clicking Manage Copy Storage as seen in Figure 9.
12
Figure 9 - Managing Copy Storage
Some of the key points to consider while enabling v irtual provisioning pools for VMAX2 array s are listed below:
If there is a pre-existing VP Snap session f or a particular application AppSync is protecting, then that pool must be added to
AppSy nc in order to prov ision new dev ices.
The Meta dev ice configurations must match, as when creating copies outside of AppSync, so the pool that is selected must also
match the RAID type of the source.
Ensure there is enough space in the pool to accommodate the number of copies. AppSync requires an additional dev ice (n+1) per
copy . If planning on creating one copy , for example, ensure there is enough space and dev ice count for two; for every source
dev ice. AppSy nc ensures there is always at least one copy available, alleviating the concern that a f ailure in production will occur
during the copy ref resh process, or the copy process does not complete successfully.
As seen in Figure 10, AppSync discovers all thin pools av ailable on the VMAX2 array , displays them in the wizard, and allows one or
more to be conf igured. If a pools consumed percentage is beyond the threshold limit, it is no longer used by AppSync for target
prov isioning, and the job will f ail if no other pool can be utilized (see key points abov e). Pools can be configured, adjusted, or deconf igured at any time, by using the option seen in Figure 9.
Should a storage pool need to be de-conf igured, an option to terminate the clone sessions AppSync created is presented, as seen in
Figure 11.
13
14
SNAPVX LINKING
Target dev ices, as observed when working with VMAX2 and TimeFinder technology , are not necessary when creating copies with
serv ice plan workf lows during the create copy phase. The copy becomes linked to a target device only when accessed, such as when
mounted. If storage groups are configured with pre-configured, and av ailable, devices, AppSync uses those devices, otherwise, if
storage groups are not utilized, AppSy nc creates new target dev ices from the default SRP. When devices are unmounted, they are not
unlinked. Target dev ices are unlinked only when they are expired from within AppSync, though are not deleted on the array , even if
AppSy nc creates them.
st
Repurposing workf lows behav e somewhat differently than service plan workf lows. Repurposing the 1 generation copy creates a
snapshot of the dev ice, also linking to a device, regardless if that device is set to mount. This differs from a service plan copy phase in
st
that a target dev ice is required f or 1 generation copies. If storage groups are utilized, and no target dev ice is available, the
st
repurposing job f ails. Refreshing the 1 generation copy will create a new snapshot and relink to the same target dev ice. Mounting
will not perf orm the linking as would be the case when working with serv ice plans. Unmounting the device does not unlink the device,
expiring the dev ice does, just as a service plan. Service plans and Repurposing workflows are described in more detail, in the AppSync
User and Administration Guide.
nd
st
nd
The 2 generation copy is created from the 1 generation copy, though differs in behavior. The 2 generation copy will not link to a
st
target during the copy phase, as does the 1 generation copy. This process is more like service plan workf lows, in that target devices
are linked upon the need to access the data, such as through a mount operation. Refreshing the copy will create a new snapshot and
relink only if the snapshot had already been linked, such as if it had been mounted. Expiring the copy, again, will unlink the dev ice and
delete the snapshot.
Some f requently asked questions (FAQs) regarding SnapVX and Linking are below:
1.
Q: When AppSync mounts VMAX3 copies, it links them to target devices. Does it unlink the target device when un-mounting?
A: No, it does not unlink when un-mounting, only when expiring.
2.
Q: Does it not make more sense to have AppSync unlink when un-mounting?
A: This is a feature of SnapVX which AppSync utilizes to increase efficiency during the next mount operation of the same SnapVX snapshot. The
device maintains the same address in the masking view settings, specifically the port group, so behaves much more efficiently.
3.
4.
Q: Can Appsync use the same targets for the next mount operation?
A: Yes, when refreshing, AppSync does an unlink and relink internally, which is transparent to the end user.
A Bronze plan f or creating local copies (local array based copies and RecoverPoint local bookmarks)
A Silver plan f or creating remote copies (remote array based copies with SRDF integration or RecoverPoint remote bookmarks)
A Gold plan f or creating local and remote copies (Local and remote bookmarks gold plans do not support SRDF)
Applications can be subscribed to dif ferent service plans based on their required SLA agreements, and/or recovery point objectives
(RPO). Serv ice plans are composed of several phases which def ine the copy schedule/on-demand, copy technology type, copy
retention, application backup type, mount options, and any other options that may be required. The dif ferent phases are depicted in
Figure 16.
16
17
AppSy nc uses the SMI-S prov ider to map the world wide names (WWNs) of the source LUNs to a corresponding VMAX array to
obtain their details. These details include the ty pe of disk, RAID type, etc. This occurs every time a service plan is run, and is not
delectable, so if there are any changes to the device on the array , AppSync is updated of those changes, such as if the size is
changed.
SMI-S prov iders should be kept online, otherwise delay s may be experienced. If an SMI-S prov ider host is taken down f or
maintenance, f or example, it should be removed from within AppSync, or the pref erred list should be modified.
For Silv er serv ice plans, AppSync also tries to find the remote device, i.e.R2, after getting the R1 details from the SMI-S Prov ider.
Both local and remote array s must be added to AppSy nc.
A FFINITIZATION RULES
Once the mapping phase completes, AppSy nc affintizes the applications based on certain rules allowing group applications to be
replicated together, i.e. when multiple applications are subscribed to the same service plan. Affinitization rules are application-specific,
as well as storage-specif ic. For example, each application has rules specific to that application, eg. SQL, Oracle, Vmware, as well as
rules specif ic to a storage system on which the application data resides. In the case where the application data resides on a VMAX
storage sy stem, the following affintization rules apply:
o
Affinitize by VMAX array ID: This rule is based on the VMAX ID (serial Number). If two databases, datastores, etc., are
on the same host but are on two dif f erent VMAX sy stems, AppSync separates them into different point-in-time copies.
Affinitize by VMAX device type: For example, thin pool dev ice vs. thick (RAID devices). Multiple databases on the same
VMAX array spread across thin and thick LUNs are split based on the type of underlying device. All applications on thin
dev ices, for example, which are on the same host, will be grouped together as part of one point-in-time copy.
Affinitize by RA Group (only for silver plans using SRDF): As part of the silver service plan execution, all devices
belonging to the same RA group are protected as one point-in-time copy .
AppSy nc supports TimeFinder VP Snap and TimeFinder Clone f or VMAX2 and SnapVX Snap and SnapVX Clone f or VMAX3 as
the supported copy technologies
TimeFinder Snap utilizing VDEVs are not supported on VMAX2, and instead, utilize VP Snaps.
By def ault all service plans have snapshot as the preferred storage technology type. If clones are desired, change the storage
order pref erence or deselect the other technologies, as seen in figures Figure 17and Figure 18.
18
If the source storage devices are thick LUNs, AppSync always creates TimeFinder Clone copies even if snapshot is
higher in the storage order pref erence
If the source storage devices are thin LUNs, AppSync always creates VP Snap copies as part of the service plan run
If the application data resides on a mix of thin and thick LUNs, AppSync alway s creates clone copies since it cannot
create VP Snaps f or thick dev ices
A specif ic copy rotation value in service plan settings can be configured. The def ault value is seven, whereas the schedule is every
24 hours, by def ault, thus creating seven copies, or one weeks worth of copies by default. In the case of VMAX2 env ironments,
plan to ensure there are at least n+1 target dev ices for each source; where n is the rotation value. This extra set of dev ices is
required due to AppSy nc expiring copies only after completing the next copy. This ensures there is always one copy, in case the
new copy f ails, or a corruption/disaster occurs, during the creation of the new copy .
o
This extra dev ice is not required with VMAX3, since with VMAX3, AppSy nc re-links the device upon refreshing the
mounted v olume not needing a target dev ice during the copy phase.
AppSy nc alway s uses the differential copy option. This is done so that AppSync does not need to perform a full synchronization in
the ev ent that a clone copy is rotated (expired from AppSync and new clone created at dif ferential point-in-time).
VMAX2 allows a maximum of eight clone copies per source device. VP Snap restores require one differential TF clone session.
Since AppSy nc creates a differential clone copy and also supports VP Snap copies for same source device, it is alway s advisable
to limit the number of clone copies to a maximum of six in the create copy phase rotation value setting.
When considering creating remote copies, such as taking snaps and clones off of the R2:
o
AppSy nc only supports SRDF/S and SRDF/A please refer to the support matrix for the current list of what
applications are supported f or each technology, e.g. supporting SRDF/A with Oracle is currently the only supported
application with SRDF/A.
For SRDF/A the link state should be in Consistent or Sy nchronized f or replications to work properly. For SRDF/S,
link state should be Sy nchronized.
Use the Silver serv ice plan to create copies off the R2. AppSync does not support Gold service plans for SRDF at
the time of publication.
Creating copies of f R2 with asynchronous mode (SRDF/A), the following should be enabled:
Dev ice lev el write pacing should be enabled please ref er to the SRDF product guide f or more details
All the dev ices which constitute a single application should belong to the same RA Group
If the source device has pre-existing session with target devices, and are configured for AppSync, those devices receive the highest priority to be
used by AppSync
2.
In case the previous rule does not yield a target device, AppSync queries the SMI-S provider in order to find information about the virtual
provisioned pool VP Snap copies. Appsync uses this pool information to get a suitable virtual provisioned pool from the list of all enabled thin
pools. AppSync stores this suitable pool information in memory so that if no suitable target devices are found, AppSync can provision new targets
for the source LUNs.
3.
When AppSync queries its database, the devices are prioritized by the following specifications:
19
If AppSync does not find target devices, and there are no valid pools configured, AppSync considers target devices which had a pre-existing
session, but is no longer being used for any copy session. It terminates the earlier session and re-uses the targets for the new copy.
5.
If a source LUN is a META LUN, AppSync searches for target META LUNs matching the same size as the source and same type as source META
i.e. stripped or /concatenated. In a situation where the target meta size and type matches, but the number of members are different, AppSync does
not consider that target as eligible.
6.
While provisioning META targets, AppSync requires a virtual provisioned pool protection level i.e. RAID type, that matches the RAID type of source
meta device. For example, if the source META LUN is at RAID-5 device, AppSync requires a thin provisioned pool with RAID-5 to provision a target
META LUN for this source LUN.
7.
If both storage group and storage pool for target and source is a META thin LUN, make sure META targets match the RAID level of META source
LUNs in the configured storage group, otherwise AppSync may not be able pick an appropriate pool for provisioning new META targets (since it
matches the RAID level wile selecting the pool). This restriction is not applicable if pools are not configured for provisioning new META LUNs.
*NOTE: For VMAX3 target device selection and provisioning, see the Storage Configuration and VMAX3 SnapVX sections.
bef ore activ ating the first point-in-time copy, the second point-in-time copy doesnt need to wait, as it is just an update of changed
tracks.
EXPIRING COPIES
Expiring copies with AppSy nc, on a VMAX array , does not delete the dev ices on the array , even if provisioned by AppSync. The
AppSy nc expire command does not equate to symsnap, symclone, or symsnapvx terminate. AppSync will re-use the target devices
f or the next snapshot, or clone, of the same source device. This reduces the time it takes to provision devices as targets. Optionally run
SY MCLI commands to validate, for example symdev show {target ID}, symsnap list, symclone list, or symsnapvx list.
For virtual machines, AppSync performs mounts by masking the copies to the ESX host, and then adds the devices to the virtual machine.
2.
In case the host is connected to VMAX via multiple masking views, and the host is not an ESX host in a cluster (for virtual or RDM mounts),
AppSync gives first preference to a masking view dedicated for that host. For example, the initiator group connected to the masking view has only
initiators for that host). In this scenario AppSync adds the target devices to that storage group; connected to the dedicated masking view.
21
3.
If AppSync doesn't find a dedicated storage group, it creates a list of storage groups with all storage groups attached to each of the masking devices
that are connected to the host and then picks up the first storage group. If the selected storage group is dedicated for GK devices (i.e. has only
gatekeeper devices), AppSync picks up the next storage group in a sorted list. If no other storage group exists except the one with only GK devices,
AppSync selects that storage group.
4.
For an ESX host in a cluster, AppSync tries to find a masking view which has connectivity to the maximum number of nodes of the ESX cluster. If it
finds such a common masking view, it uses the storage group to add the devices during the mount phase. In case any ESX nodes in cluster are not
connected to this common masking view, AppSync searches for a masking view for that node as outlined in step 2 and 3. If no storage group for
any node of the cluster is found, the mount operation fails with appropriate exception.
5.
When the selected masking view is connected to a cascaded storage group i.e. storage group containing a list of other storage groups, AppSync
sorts the storage group by its name and picks up the first storage group checking the storage group is not dedicated for GK devices. If the selected
storage group is dedicated for GK devices, AppSync picks the next storage group in sorted list.
6.
When working with SQL cluster mounts, if the use dedicatedStorageGroup option is selected (selected by default), and if AppSync is unable to
find any dedicated storage group for the mount host, an exception is displayed and the mount fails.
7.
During unmount, AppSync finds the storage group using the same rules as described previously and removes the devices from the storage group
as shown in Figure 21.
22
REPURPOSING WORKFLOWS
Repurposing workf lows are v ery similar to the service plan workflows, in that the same phases are followed, as seen in Figure 16.
There are dif f erences, howev er, such as the repurposing workflow allows taking multi-generation copies, and also, offers the ability to
take array based copies off of a RecoverPoint bookmark (placing RecoverPoint bookmarks in image access mode as a temporary
st
placeholder in order to create an array based copy ). The repurposing workf low offers a 1 generation copy which can be used f or
nd
restore purposes, since it can be used as a gold copy, and also serves as a source for multiple 2 generation copies. Both
st
generation copies can be mounted and written to, but only the 1 can be restored to production.
st
nd
1 Generation Copies: Can be used f or restore purposes and serves as the same source for multiple 2 generation copies
2 Generation Copies: Cannot be used f or restore purposes, and is created from a 1 generation copy (e.g. snap-of-snap)
nd
st
Repurposing also dif f ers in the way target devices are managed in a VMAX3 env ironment. When working with VMAX3 array s and
serv ice plan workf lows, devices are only required when linking, as occurs when the dev ices are mounted. With the repurposing
st
nd
workf low, howev er, the 1 generation copy automatically links to devices, regardless if the devices are mounted. The 2 generation
copy shares the same rules as service plan copies, where linking only occurs during the mount operation. With VMAX2, howev er, the
repurposing workf low behav es the same as the service plan workf low, as target devices are alway s required. With this knowledge,
ensure that there are either target dev ices available in the storage group AppSy nc is configured with, or no storage group is configured,
allowing AppSy nc to create dev ices as necessary.
st
Lastly , repurposing 1 generation copies do not require the additional target devices, as do service plans. Using Service Plans with
VMAX2 and TimeFinder, one additional target dev ice per source is required. The prev ious copy is expired only after the new copy is
f ully complete. This means that VMAX2 TimeFinder serv ice plans require an N+1 copy count. That is to say, AppSync requires one
additional target dev ice f or every source device, for every service plan job. For example, a SQL database is set to a copy expiration
count of sev en (default expiration count), so when working with VMAX2 and TimeFinder, there needs to be eight copy targets available
in a storage group, or av ailable f or AppSy nc to create. With the repurposing workflow, however, AppSync does not require the N+1
rotation count, as it will reuse the same target dev ices. When working with VMAX3 SnapVX copies, repurposing workf lows are similar
to VMAX2, in that only one dev ice is required, and does not require the N+1 copy count.
Repurposing gen 1:
o
Create copy phase will create the copy of the device and link it to a target
Refresh will create a new copy and relink to the same target dev ice
Unmount will not unlink the target dev ice, and only unmounts from the host (speeds up re-mounting)
Expire will unlink the target dev ice and delete the copy (any target device created remains on the array)
Repurposing gen 2:
o
Create copy phase will create the copy , howev er, will not be linked to a target dev ice
23
When the copy needs to be accessed, i.e. mounted, it is linked to a target device, then mounted
Refresh will create a new copy , and relinked if the original copy was already linked to a target device
Expire will unlink the target dev ice, if linked, and deleted (any target device created remains on the array )
REPURPOSING USE CA SE
The f ollowing depicts a common repurposing workf low use case. A database developing group wishes to create multiple copies of
production. Users require a point-in-time copy to serves as the baseline copy, the gold copy, for the development groups testing.
st
Since multiple users need the same PIT copy , they must use a common sourced copy, from the same PIT. This is the 1 generation
copy , also commonly referred to as a gold copy . This copy, by default, is application consistent.
nd
Each user has the ability to create 2 generation copies, presenting them to different hosts. Each copy is autonomous, and can be
st
ref reshed upon demand, or schedule. The 1 generation copy can be ref reshed as well, independently, thus providing a div erse work
st
env ironment. The 1 copy , being that it is not mounted or changed, can also serve as a restorable copy.
The f ollowing expert of the process log depicts a common flow for creating a 1 generation copy, for the first time, with a mounting
option.
Phase
Master Phase
Application discovery
Application discovery
Application mapping
Application mapping
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Create 1st gen copy
Unmount previous
copy
Unmount previous
copy
Mount copy
Mount copy
Message
Beginning execution of service plan
Application discovery phase beginning
Application discovery phase completed successfully
Application mapping phase beginning
Application mapping phase completed successfully
Create 1st gen copy phase beginning
Skipping unmount because there were no previously mounted copies found for
the applications under protection during this cycle.
Attempt number 1 to create a VSS copy of the application.
Starting backup of SQL Server databases.
VSS application freeze succeeded on host amssqlp1.
Backup of database completed successfully.
VSS application thaw succeeded on host amssqlp1.
The SnapVX snapshot for the source device 003B8 on the array is
EMC_SYNC_ASPECT1474905607970.
The SnapVX snapshot for the source device 003B9 on the array is
EMC_SYNC_ASPECT1474905607970.
Refreshing the SnapVX replication relationships in storage array.
The SnapVX snapshot EMC_SYNC_ASPECT1474905607970 for the source device
003B8 on the array is linked in nocopy mode with target device 003BC.
The SnapVX snapshot EMC_SYNC_ASPECT1474905607970 for the source device
003B9 on the array is linked in nocopy mode with target device 003BD.
Create 1st gen copy phase completed successfully
Skipping unmount because there were no previously mounted copies found for
the applications under protection during this cycle.
Unmount previous copy phase for applications completed successfully
Mount copy phase beginning
Beginning mount on mount host
24
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Mount copy
Recover copy
Recover copy
Recover copy
25
Since AppSync supports TimeFinder VP Snap (thin pools), not TimeFinder Snap (using VDEVs), while restoring a VP Snap snapshot, VMAX
consumes one TimeFinder clone session to perform the restore operation. It is very important to ensure AppSync is configured to make no more
than six copies of the same application, when also creating snaps, due to this restriction.
2.
Once the restore completes, AppSync splits the pair in order to perform a restore from the same copy again. AppSync can also refresh the same
copy (in case it is expired from AppSync as part of rotation), performing a differential copy during next service plan run.
3.
At the end of restore from VP Snap copies, VMAX creates one additional session in "RESTORED" state for the same source-target pair. AppSync
terminates this RESTORED session at the end of restore, so that the same copy can be used again for restore as well as restore from other copy. If
the session is left in the RESTORED state, it cannot restore from other copies.
TROUBLESHOTTING TIPS
AppSy nc uses SMI-S prov ider to manage VMAX storage sy stem and many times failure of an AppSync service plan and other
operations related to VMAX array s are due to SMI-S prov ider issues. In such cases, it is always recommended to collect log files
related to SMI-S prov ider. Below is a list of log f iles which need to be collected from SMI-S provider host :
1.
Stop ecom
Start ecom on windows c:\program files\emc\ecim\ecom\bin\sm_ser vice start ecom.exe on Linux from the directory /opt/emc/ecim/ecom/bin
./ECOM d
6.
Wait for ecom to fully start use the testsmiporvider dv command to verify ecom is fully started.
7.
Instead of launching ecom as a service on Windows or a background process on Linux , run it from the command line to collect Solutions Enabler
traces. To do that, open a command prompt on Windows or Linux.
2.
3.
4.
Re-start ecom from the command line on Windows by typing ecom and hitting enter, or on Linux, type in ./ECOM and hit enter.
5.
Wait for ecom to fully start, and use the testsmiporvider dv command to verify ecom is fully started.
6.
Issue: A copy fails after one of the devices provisioned by AppSync was deleted outside of AppSync
Troubleshooting steps: Never delete devices which are being used by AppSync as targets.
2.
Issue: Discovery of VMAX systems using the SMI-S Provider hangs or is very slow
Troubleshooting steps: Run discovery outside of AppSync using TestSmisProvider.exe to determine if it is also behaving in a similar way.
Check if there are arrays which the SMI-S provider is discovering, that are not correctly configured on host, do not have the necessary gatekeepers,
or redundant to the SMI-S host. In such case, remove those arrays from automatic discovery using SYMAVOID file as described in Procedure to
prevent Automatic Discovery of Arrays section of this whitepaper.
3.
27
Troubleshooting steps: From the symdev show cli command, verify that the link is not in the consistent state. Per the AppSync User and
Administration Guide, ensure the R1>R2 link state is Synchronized for SRDF/S or Consistent for SRDF/A.
4.
Issue: Clone replication fails with error pointing to exceeding number of sessions
Troubleshooting steps: TimeFinder is limited to eight differential pairing sessions. AppSync utilizes differential sessions, and offers no more than
seven copies, since one session must be reserved for restoring VP Snap sessions. In addition, since an extra copy is needed for rotation (N+1)
AppSync does not delete the prior copy till the next copy is complete, the maximum number of clone copies should be set to six. Please refer to the
Copy session limits section in the AppSync User and Administration Guide, for more details.
5.
Issue: Service Plan is configured with an expiration value of 1 but AppSync is creating two clone sessions
Troubleshooting steps: AppSync always maintains two clone sessions minimum, due to the fact that AppSync ensures a copy is fully created
before expiring the previous copy. This method is considered an N+1 copy count, and cannot be changed.
6.
https://elabnavigator.emc.com/eln/modernHomeDataProtection
7.
8.
CONCLUSION
In conclusion, this whitepaper explains key information and concepts when protecting an application residing on VMAX storage using
AppSync. It discusses the inner operations of AppSync software while protecting applications on VMAX2 using TimeFinder VP Snap
and Clone, and VMAX3 using SnapVX Snap and Clone copies along with deployment and troubleshooting tips. The main differences
between deploying in a VMAX2 and VMAX3 environment is also identified.
28