Resharding redis cluster

[AntonArsentiev]

Hello everyone.

I'm using your library version 6.4.0.

I want to reshard redis cluster to update the redis version on each node of the cluster.

After transferring slots to neighboring cluster nodes and updating redis on an empty node, I encountered this situation:

cluster client receives response "ASK host:port" and then program execution crashes with an error

await target_node.execute_command("ASKING")
AttributeError: 'NoneType' object has no attribute 'execute_command'

because cluster client no longer knows anything about old node after reinitialization, since there are no slots on old node anymore.

Please tell me how I can reshard without this error, or maybe I should use a newer version of the library, where client will know about node, even if there are no slots on it.

[AntonArsentiev]

If I understood correctly, situation will be similar if I want to expand the cluster by adding another node and transferring some slots to it. Please correct me if I'm wrong.

Please tell me how to perform reshard correctly in my situation.

[arlosnow1231-stack]

This happens because the client drops nodes that have no slots, so it can’t handle ASK redirects to them. As a workaround, try refreshing the cluster topology (or recreating the client) during resharding, or avoid leaving a node with zero slots. Upgrading may also help.

…

On Tue, Mar 31, 2026, 6:25 AM AntonArsentiev ***@***.***> wrote: *AntonArsentiev* left a comment (redis/redis-py#4015) <#4015 (comment)> If I understood correctly, situation will be similar if I want to expand the cluster by adding another node and transferring some slots to it. Please correct me if I'm wrong. Please tell me how to perform reshard correctly in my situation. — Reply to this email directly, view it on GitHub <#4015?email_source=notifications&email_token=CAXLFZUTIRFOG2XH2MGFLUL4TOTL5A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJWGE4TGMJZG4YKM4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4161931970>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CAXLFZTFHONOXVQX5XTMIML4TOTL5AVCNFSM6AAAAACXGNYXW6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNRRHEZTCOJXGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[petyaslavova]

Hi @AntonArsentiev, can you try with the latest release? There were several improvements throughout the cluster topology update logic in the last few months.

[AntonArsentiev]

Hi @petyaslavova, I've looked at the code of the latest version of the library and it seems I'm catching the same error and here's why

1. After these lines self.nodes_cache will contain only those nodes that have slots, although initially another master node and its replica will be indicated in self.startup_nodes.
   https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1951-L1955
2. If I do reshard, I can get an ASK on these lines
   https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1162-L1174
3. There is no topology update in step 2, as shown in these lines
   https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1132-L1161
4. Then on the next line I will get target_node=None and then I'll get the error I mentioned earlier.
   https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1044

[nadiatazi262-tech]

Hi @AntonArsentiev, thanks for the detailed breakdown — that makes sense. It does look like the topology isn’t getting refreshed after the ASK, which could explain the missing target node. Let me double-check this flow on the latest version and see if we’re missing an update step there

…

On Wed, Apr 1, 2026, 5:48 AM AntonArsentiev ***@***.***> wrote: *AntonArsentiev* left a comment (redis/redis-py#4015) <#4015 (comment)> Hi @petyaslavova <https://github.com/petyaslavova>, I've looked at the code of the latest version of the library and it seems I'm catching the same error and here's why 1. After these lines self.nodes_cache will contain only those nodes that have slots, although initially another master node and its replica will be indicated in self.startup_nodes. https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1951-L1955 2. If I do reshard, I can get an ASK on these lines https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1162-L1174 3. There is no topology update in step 2, as shown in these lines https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1132-L1161 4. Then on the next line I will get target_node=None and then I'll get the error I mentioned earlier. https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1044 — Reply to this email directly, view it on GitHub <#4015?email_source=notifications&email_token=B7HY7376E6CCFP2MJH52XJL4TTXYRA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJWHEYTSNJUGM42M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4169195439>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/B7HY735GAFHDIBR7ZOADGC34TTXYRAVCNFSM6AAAAACXGNYXW6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNRZGE4TKNBTHE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[arlosnow1231-stack]

Alright, keep me posted. On Wed, Apr 1, 2026, 12:59 PM nadiatazi262-tech ***@***.***> wrote:

…

*nadiatazi262-tech* left a comment (redis/redis-py#4015) <#4015 (comment)> Hi @AntonArsentiev, thanks for the detailed breakdown — that makes sense. It does look like the topology isn’t getting refreshed after the ASK, which could explain the missing target node. Let me double-check this flow on the latest version and see if we’re missing an update step there On Wed, Apr 1, 2026, 5:48 AM AntonArsentiev ***@***.***> wrote: > *AntonArsentiev* left a comment (redis/redis-py#4015) > <#4015 (comment)> > > Hi @petyaslavova <https://github.com/petyaslavova>, I've looked at the > code of the latest version of the library and it seems I'm catching the > same error and here's why > > 1. After these lines self.nodes_cache will contain only those nodes > that have slots, although initially another master node and its replica > will be indicated in self.startup_nodes. > > https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1951-L1955 > 2. If I do reshard, I can get an ASK on these lines > > https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1162-L1174 > 3. There is no topology update in step 2, as shown in these lines > > https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1132-L1161 > 4. Then on the next line I will get target_node=None and then I'll get > the error I mentioned earlier. > > https://github.com/redis/redis-py/blob/master/redis/asyncio/cluster.py#L1044 > > — > Reply to this email directly, view it on GitHub > < #4015?email_source=notifications&email_token=B7HY7376E6CCFP2MJH52XJL4TTXYRA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJWHEYTSNJUGM42M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4169195439>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/B7HY735GAFHDIBR7ZOADGC34TTXYRAVCNFSM6AAAAACXGNYXW6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNRZGE4TKNBTHE> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > — Reply to this email directly, view it on GitHub <#4015?email_source=notifications&email_token=CAXLFZRN7YTXO3LFHNNBRLD4TVKHZA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJXGE4TMMJSHE42M4TFMFZW63VHMNXW23LFNZ2KKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4171961299>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CAXLFZR55M5QYILZ3UM6BBD4TVKHZAVCNFSM6AAAAACXGNYXW6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNZRHE3DCMRZHE> . You are receiving this because you commented.Message ID: ***@***.***>