Problemsby GlyphRedis upgrades break working setups
9 posts from 9 people8d activeSteady, +1 this week
Posts per day
The posts behind it
Atomic unread messages update for member in group chat
I have a group chat with members, when a user (member) opens a group I set users_opened_chats:<user_id> as group_id in redis and set unread_message_count to 0 at the member's row in database, so I can easily check what members are now online in this group and using this...
[Bug]: Daily tag spend is lost after Redis dequeue if writer exits before DB commit
### Description A daily Tag spend batch that has already been accepted into the Redis buffer can disappear if the writer process is killed after dequeue but before PostgreSQL commit. I reproduced this with two independent writer processes sharing real PostgreSQL and Redis. After...
prefect-redis: Publisher silently drops a message published while a flush awaits the dedup check
### Bug summary `prefect_redis.messaging.Publisher._publish_current_batch` loses any message that `publish_data` appends while a flush is waiting on the deduplication round-trip. The flush passes `self._batch` to `Cache.without_duplicates`, which builds its `SET NX` pipeline...
Inactive tag concurrency limit still creates a lease per task, and a 410 on renewal logs ERROR despite suppress_warnings
...forces the `410`, so it shows what happens after a lease is gone, not why it goes. **Workaround:** we've deleted the two inactive limits, so those tags no longer take the lease path. It's too early to confirm from our Sentry data that this stopped the errors. **Related:** #23658...
Deployment concurrency lease renewal gets 410 after the server applies the terminal state, and crashes a run that completed
### Bug summary A deployment-concurrency lease renewal can get `410 Gone` after the server has applied the run's terminal state and released the lease, while the client has not yet processed the answer to that terminal proposal. The renewal thread then spends its three attempts...
issue: instance stops accepting websocket connections until restart after its Redis connection drops
...without `retry_on_timeout` | connect finishes after 1.5 s | **Additional Information** Workaround today: remove `retry_on_timeout` from `WEBSOCKET_REDIS_OPTIONS` on every instance. A rolling restart is enough. A dropped Redis connection then recovers after about 1.5 s. An...
[BUG] A2A task cache passes a percent-encoded Redis password from REDIS_URL
### Description `_parse_redis_url` in `crewai/a2a/utils/task.py` builds the aiocache Redis config from `REDIS_URL` with `urlparse`. `urlparse` does not decode percent-encoding, and the function passes `parsed.password` to the config unchanged. A password that contains `@`, `/`,...
[Bug]: DualCache writes Redis entries with `default_in_memory_ttl`, so a configured `default_redis_ttl` never takes effect
### Description `DualCache` accepts a Redis TTL in three places and never reads it back. `self.default_redis_ttl` is assigned at `litellm/caching/dual_cache.py:77` (constructor), `:84` (`update_cache_ttl`) and `:106` (`attach_redis_cache`), and there is no read of it anywhere in...
Explicit falsy Redis client overrides should not fall back to settings
...`0` values for `port`, `db`, and `health_check_interval`. ### Reproduction This makes it impossible to explicitly disable TLS for a local/test client while inheriting a shared settings object, and it also prevents callers from selecting database 0 when settings select another...
Companies and products named
About this problem
Evidence
9 posts from 9 people in 5 places, about 1 a week over 73 days. Mostly on PrefectHQ/prefect, BerriAI/litellm, crewAIInc/crewAI. Tools named alongside: Prefect, Docker, PostgreSQL, Claude Code.
Frustration Frustration 1 of 3· Seen on GitHub issues, Stack Overflow
How it was grouped
Posts that state a pain and match the "breaking-changes" rule about Redis. First post Aug 29, 2026, latest Nov 9, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
The same problem elsewhere
GitHub upgrades break working setups
97 posts from 83 people on GitHub issues, Hacker News, Stack Overflow.
Upgrades and new releases break working setups
81 posts from 65 people on GitHub issues, Reddit, App Store reviews.
Claude Code upgrades break working setups
71 posts from 64 people on GitHub issues, Hacker News, Reddit.
Codex upgrades break working setups
68 posts from 61 people on GitHub issues, Reddit.
DuckDB upgrades break working setups
45 posts from 35 people on GitHub issues.
Claude upgrades break working setups
36 posts from 34 people on GitHub issues, Reddit, Hacker News.
ClickHouse upgrades break working setups
31 posts from 20 people on GitHub issues.
Anthropic upgrades break working setups
27 posts from 22 people on GitHub issues.
ChatGPT upgrades break working setups
27 posts from 27 people on GitHub issues, Community forums, App Store reviews.
PyTorch upgrades break working setups
26 posts from 21 people on GitHub issues, Reddit.
OpenAI upgrades break working setups
26 posts from 26 people on GitHub issues, Reddit, Community forums.
Open WebUI upgrades break working setups
25 posts from 2 people on GitHub issues.
Other Redis problems
Redis rate limits and usage caps block work
6 posts from 6 people on GitHub issues.
Redis users keep filing bugs
Top post: "7.0.0rc2 Flight SQL/ADBC validation: source-install dependency failures and metadata gaps", one of 3 posts from 1 person on GitHub issues.
History