highperfocused/danswer
1
0
Fork 0
mirror of https://github.com/danswer-ai/danswer.git synced 2025-09-29 05:15:12 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix timing calculations and don't spam the queue lengths check from every task

Browse Source
This commit is contained in:
Richard Kuo (Danswer)
2025-01-05 21:13:08 -08:00
parent 98005510ad
commit e3947e4b64
1 changed files with 57 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
backend/onyx/background/celery/tasks/vespa/tasks.py
View File

@@ -1,3 +1,4 @@
import random
import time
import traceback
from datetime import datetime
@@ -77,6 +78,7 @@ from onyx.utils.variable_functionality import (
)
from onyx.utils.variable_functionality import global_version
from onyx.utils.variable_functionality import noop_fallback
from shared_configs.configs import MULTI_TENANT
logger = setup_logger()
@@ -781,12 +783,17 @@ def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
return False
# print current queue lengths
phase_start = time.monotonic()
# we don't need every tenant polling redis for this info.
if not MULTI_TENANT or random.randint(1, 100) == 100:
r_celery = self.app.broker_connection().channel().client # type: ignore
n_celery = celery_get_queue_length("celery", r_celery)
n_indexing = celery_get_queue_length(
OnyxCeleryQueues.CONNECTOR_INDEXING, r_celery
)
n_sync = celery_get_queue_length(OnyxCeleryQueues.VESPA_METADATA_SYNC, r_celery)
n_sync = celery_get_queue_length(
OnyxCeleryQueues.VESPA_METADATA_SYNC, r_celery
)
n_deletion = celery_get_queue_length(
OnyxCeleryQueues.CONNECTOR_DELETION, r_celery
)
@@ -818,29 +825,30 @@ def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
f"external_group_sync={n_external_group_sync} "
f"permissions_upsert={n_permissions_upsert} "
)
timings["queues"] = time.monotonic() - timings["start"]
timings["queues"] = time.monotonic() - phase_start
# scan and monitor activity to completion
phase_start = time.monotonic()
lock_beat.reacquire()
if r.exists(RedisConnectorCredentialPair.get_fence_key()):
monitor_connector_taskset(r)
timings["connector"] = time.monotonic() - phase_start
timings["connector"] = time.monotonic() - timings["queues"]
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisConnectorDelete.FENCE_PREFIX + "*"):
lock_beat.reacquire()
monitor_connector_deletion_taskset(tenant_id, key_bytes, r)
timings["connector_deletion"] = time.monotonic() - timings["connector"]
timings["connector_deletion"] = time.monotonic() - phase_start
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisDocumentSet.FENCE_PREFIX + "*"):
lock_beat.reacquire()
with get_session_with_tenant(tenant_id) as db_session:
monitor_document_set_taskset(tenant_id, key_bytes, r, db_session)
timings["document_set"] = time.monotonic() - phase_start
timings["document_set"] = time.monotonic() - timings["connector_deletion"]
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisUserGroup.FENCE_PREFIX + "*"):
lock_beat.reacquire()
monitor_usergroup_taskset = fetch_versioned_implementation_with_fallback(
@@ -850,29 +858,29 @@ def monitor_vespa_sync(self: Task, tenant_id: str | None) -> bool:
)
with get_session_with_tenant(tenant_id) as db_session:
monitor_usergroup_taskset(tenant_id, key_bytes, r, db_session)
timings["usergroup"] = time.monotonic() - phase_start
timings["usergroup"] = time.monotonic() - timings["document_set"]
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisConnectorPrune.FENCE_PREFIX + "*"):
lock_beat.reacquire()
with get_session_with_tenant(tenant_id) as db_session:
monitor_ccpair_pruning_taskset(tenant_id, key_bytes, r, db_session)
timings["pruning"] = time.monotonic() - phase_start
timings["pruning"] = time.monotonic() - timings["usergroup"]
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisConnectorIndex.FENCE_PREFIX + "*"):
lock_beat.reacquire()
with get_session_with_tenant(tenant_id) as db_session:
monitor_ccpair_indexing_taskset(tenant_id, key_bytes, r, db_session)
timings["indexing"] = time.monotonic() - phase_start
timings["indexing"] = time.monotonic() - timings["pruning"]
phase_start = time.monotonic()
for key_bytes in r.scan_iter(RedisConnectorPermissionSync.FENCE_PREFIX + "*"):
lock_beat.reacquire()
with get_session_with_tenant(tenant_id) as db_session:
monitor_ccpair_permissions_taskset(tenant_id, key_bytes, r, db_session)
timings["permissions"] = time.monotonic() - timings["indexing"]
timings["permissions"] = time.monotonic() - phase_start
except SoftTimeLimitExceeded:
task_logger.info(
"Soft time limit exceeded, task is being terminated gracefully."