Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

BLPOP with a 0 timeout queued in MULTI never replies at EXEC, and leaves the key unusable for every client

Closed
  • Is something wrong?
#2,266 0 comments 0 reactions 1 assignee View on GitHub

Maintainers usually reply within 2 days

@badrishc is already working on this.

Since Oct 22, 2026.

  • #2285 by @badrishc — merged

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
55/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
csharp
Domain
backend, databases

Research direction

The issue identifies BLPOP inside MULTI/EXEC as the failing path, but names no source files or tests. Start by tracing how Garnet executes blocking list commands inside transactions and find the existing tests for empty-list BLPOP with a zero timeout. Reproduce the hang and verify that EXEC returns [nil] and a fresh connection can use the key afterward; check the other listed blocking commands for the same behavior.

Written by the indexing model from the issue text.

Description

Describe the bug

BLPOP key 0 queued inside MULTI and then run by EXEC, on an empty or missing list, never
produces a reply. The client is stuck, and so is the key: a later write to that key from a fresh
connection also never replies. Unrelated keys keep working, and FLUSHALL does not help — only
restarting the process does.

DEL k1
MULTI            -> OK
BLPOP k1 0       -> QUEUED
EXEC             -> (no reply, ever)
PING             -> (no reply; the connection is unusable)

then, from a second, brand-new connection:

PING             -> PONG            the server is alive
SET unrelated 1  -> OK              other keys are fine
LPUSH k1 y       -> (no reply, ever)

So one client that issues a queued BLPOP ... 0 permanently removes a key from service for
everybody, and burns a connection each time something touches it. Reaching it needs no special
privilege — a single ordinary transaction does it — which makes it a denial of service on any key an
untrusted client can name.

It is specific to the blocking-forever case inside a transaction. All of these are fine:

variant Garnet
MULTI; BLPOP k 1; EXEC (finite timeout) nil ✅
MULTI; LPOP k; EXEC nil ✅
BLPOP k 0 outside MULTI, list non-empty returns the element ✅
MULTI; BLPOP k 0; EXEC, list non-empty returns the element ✅
MULTI; BLPOP k 0; EXEC, list empty hangs forever ❌
Steps to reproduce the bug
  1. Start Garnet 2.2.2: docker run -d -p 6400:6379 ghcr.io/microsoft/garnet:2.2.2
  2. On one connection, from a clean keyspace (FLUSHALL):
DEL k1
MULTI
BLPOP k1 0
EXEC            <- no reply arrives; the connection is now stuck
  1. Open a second connection and try to use that key:
PING            -> PONG
LPUSH k1 y      <- no reply arrives either
  1. FLUSHALL from a third connection does not restore the key; restarting the process does.

Each row below is an independent sequence from a clean keyspace, each on its own key (sharing one
key would make the controls look broken too, since the first case wedges it). "no reply" means
nothing arrived within the client timeout.

case expected Redis 8.10.2 Kvrocks KeyDB Pika Dragonfly Garnet 2.2.2
MULTI; BLPOP k1 0; EXEC [nil] [nil] [nil] [nil] [nil] [nil] no reply
… then PING on the same connection PONG PONG PONG PONG PONG PONG no reply
MULTI; BLPOP k3 1; EXEC (finite timeout) [nil] [nil] [nil] [nil] [nil] [nil] [nil]
MULTI; LPOP k4; EXEC [nil] [nil] [nil] [nil] [nil] [nil] [nil]
BLPOP k5 0 outside MULTI, non-empty list [k5, x] [k5, x] [k5, x] [k5, x] [k5, x] [k5, x] [k5, x]
MULTI; BLPOP k6 0; EXEC, non-empty list [[k6, x]] [[k6, x]] [[k6, x]] [[k6, x]] [[k6, x]] [[k6, x]] [[k6, x]]
afterwards, fresh connection: LPUSH k1 y 1 1 1 1 1 1 no reply

A self-contained Python 3 reproducer (standard library only) that runs every case, prints this table
and then probes the wedged key from fresh connections is attached as repro.py; it exits 0 when the
bug reproduces.

Expected behavior

EXEC should reply [nil], as every other implementation here does. The BLPOP documentation
specifies this case and gives exactly the reason it must not block — it would hold up the server:

BLPOP inside a MULTI / EXEC transaction

Using BLPOP inside a MULTI / EXEC block does not make a lot of sense as it would require
blocking the entire server in order to execute the block atomically, which in turn does not allow
other clients to perform a push operation. For this reason the behavior of BLPOP inside
MULTI / EXEC when the list is empty is to return a nil multi-bulk reply, which is the same
thing that happens when the timeout is reached.

If you like science fiction, think of time flowing at infinite speed inside a MULTI / EXEC block...

The same applies to the other blocking list and sorted-set commands (BRPOP, BLMOVE,
BRPOPLPUSH, BLMPOP, BZPOPMIN, BZPOPMAX, BZMPOP) — we only verified BLPOP, but they are
worth checking together.

Screenshots

No response

Release version

v2.2.2

IDE

No response

OS version

Ubuntu 24.04.5 LTS, kernel 7.0.0-38-generic (x86_64), Docker. The container reports
os:Unix 7.0.0.38.

Additional context

Environment where the bug was reproduced contains Garnet 2.2.2, from the published container image ghcr.io/microsoft/garnet:2.2.2
(digest sha256:536185d34d4783856401f79f85d98c0cd46b417fa1a1c2ddd4441bc4012c4f34).
INFO server reports garnet_version:2.2.2. Source read at
456925d8b36eb139d0a37937b85c085ab815c482 (tag v2.2.2).

Dominant language
C#
Stars
12.1k
Forks
716
Avg merge
2d 2h
Merged PRs (30d)
65

Getting set up

  • Ships a Dockerfile or Docker Compose file
  • No pull request template
  • No contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from microsoft/garnet

All issues in microsoft/garnet

Similar issues

More C# issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.