Dell Unity: How Asynchronous Replication and Space Reclamation Affect Array Performance

Summary: This article explores the impact of asynchronous replication on array performance during space reclamation from large virtual machines (VM).

This article applies to This article does not apply to This article is not tied to any specific product. Not all product versions are identified in this article.

Symptoms

Massive latency throughout ALL storage resources (entire array) can be seen when reproducing the following steps:
 
1. Virtual Machines (VM) on Host have no usage, and are offline - no I/O - fully idle - (meant to be deleted).
2. VM is being deleted from Host (Space Reclamation).
3. Everything is working fine. All LUNs and Hosts in the Array have expected Response Times.
4. [Trigger] Replication starts either manually or by the Recover Point Objective (RPO). Latency increases in ALL LUNs and Hosts - Data became Unavailable.

Cause

Asynchronous Replication:
- When the Replication Session is paused, the Reclamation Rate does not stop nor slow down.
- While the Replication Session is paused, the work accumulates on the array.
- Once the Replication Session is resumed (either manually or automatically), this pushes the accumulated work, from when it was paused.
- Depending on the amount of work accumulated, this might cause latency, in this case, severe.
- This is Function As Design.

Space Reclamation:
- As per Best Practices, 64 TB it is not recommended for ESXi block LUNs, specially combined with Asynchronous Replication. 
- When the Array queries through 40 TB of empty space to find blocks with Variable-length map-for-read, this is a lot of no-value-add work.
- While that happens, the load increases and impacts both LUN and Array (all resources).

Resolution

To avoid massive latency when these conditions are met, it is advised to:

1. Stop and Delete the Replication Session.
2. Purge the unused thin and uncommitted space (with data migration) to a smaller LUN. 
3. Create a new Replication session from the smaller LUN, instead.
 

Alternatively, if the Migration Datastore is obsolete, and it is meant to be removed, our recommendations are:

1. Delete the Replication Session.
2. Delete affected LUN.
3. Create a smaller LUN in the future, following Best Practices. 


Further considerations: 
- Misconfiguring LUNs, ESXi hosts/VMs, and replication can have unclear consequences, even with best practices in place.
- For this case in particular, this is what was causing latency at this massive scale.
- According to Development, this is an expected behavior, and Unity is Functioning As Designed.
 

Reference Documentation:
- Dell EMC Unity: VMware vSphere Best Practices 
- vSphere Storage Appliance (Community - External Link)

Additional Information

To confirm this is the same behavior, verify that Replication is or was being triggered, look at the following files live on the array:

Live on Array:
- /EMC/backend/log/shared/EMCSystemLogFile.log
/EMC/C4Core/log/c4_safe_ktrace.log


 

Manually triggered by the Admin user:
14:10:21 service@spb:~/user# tailf 00_emc_backend_log_shared/EMCSystemLogFile.log | grep -i "refresh the replication"
B       05/04/21 14:10:48.405 ReplicPlugin       610021 [INFO] Audit: User Admin started to refresh the replication session rep_sess_res_xx_res_xx_CKM00xxxxxxxxx_CKM00xxxxxxxxx.


Automatically, by Recovery Point Objective (RPO):
B       05/04/21 14:45:33.687 std              ddf1a70d c4_safe_ktrace   SASPMC 0 (BE99) API INFO set_dynamic_int_coalescing: nr_busy_ports=1 ict=100 icc=16
B       05/04/21 14:45:33.708 sade              9f3f70f c4_safe_ktrace   DIC: 6:[SVDM_B] Handle request /dic_s/DicXmlAsyncMsgService/async=4=302002=503370607
B       05/04/21 14:45:33.708 sade              9f3f70f c4_safe_ktrace   DIC: 6:[SVDM_B] Send back /dic_s/DicXmlAsyncMsgService/async=4=302002=503370607 response (200 OK)
B       05/04/21 14:45:33.708 sade              9f3f70f c4_safe_ktrace   DIC: 6:[SVDM_B] Response for /dic_s/DicXmlAsyncMsgService/async=4=302002=503370607 sent
B       05/04/21 14:45:33.708 sade             e5243708 c4_safe_ktrace   XML: 6:[core] 
B       05/04/21 14:52:07.605 mlu               16d001c [INFO] System: Snapshot 81604378643_CKM00201901039_0000.ckpt000_3275157210146496.2_684 has been created for LUN 23.
B       05/04/21 14:52:07.710 mlu               12d0003 [INFO] System: Operation Create Replica started by 900002a4a on 9000000fd.
B       05/04/21 14:52:07.724 mlu               16d0027 [INFO] System: The new name for snapshot being destroyed is Destroying_20210106115207.694+00-000.
B       05/04/21 14:52:07.738 mlu               16d0027 [INFO] System: The new name for snapshot being destroyed is Destroying_20210106115207.694+00-001.
 

Note: Every time the RPO is reached or a manual update is started, a snapshot refresh occurs. A single snapshot refresh can be considered a destroy or re-create.

Affected Products

Dell EMC Unity, Dell EMC Unity Family |Dell EMC Unity All Flash, Dell EMC Unity Hybrid
Article Properties
Article Number: 000184863
Article Type: Solution
Last Modified: 21 Oct 2025
Version:  5
Find answers to your questions from other Dell users
Support Services
Check if your device is covered by Support Services.