phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-03 01:48:05 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(sqlstore): increase connection pool size to prevent exhaustion

Browse source 
Increases SQLAlchemy connection pool from default 5+10 to 20+30 to handle
  concurrent API operations and tests. Prevents deadlocks in conversation API
  integration tests when using Postgres, where multiple concurrent requests
  exhaust the small default pool.

  Also adds pool_recycle=3600 to prevent stale connections.

Signed-off-by: Charlie Doern <cdoern@redhat.com>
This commit is contained in:
Charlie Doern 2025-11-12 11:42:29 -05:00
parent 63137f9af1
commit 86ed32817a
1 changed files with 6 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/llama_stack/providers/utils/sqlstore/sqlalchemy_sqlstore.py
View file

@ -84,9 +84,15 @@ class SqlAlchemySqlStoreImpl(SqlStore):
connect_args["timeout"] = 5.0
connect_args["check_same_thread"] = False # Allow usage across asyncio tasks
# Configure connection pool for concurrent access
# Default pool_size=5 + max_overflow=10 = 15 total connections
# Increase to handle concurrent tests and API operations (especially with Postgres)
engine = create_async_engine(
self.config.engine_str,
pool_pre_ping=True,
pool_size=20, # Increased from default 5
max_overflow=30, # Increased from default 10
pool_recycle=3600, # Recycle connections after 1 hour to prevent stale connections
connect_args=connect_args,
)