fight various timeouts at high tenant count

This commit is contained in:
Christian Schwarz
2024-01-09 09:45:37 +00:00
parent 89fe8cb60b
commit 5f4fe8f72a
2 changed files with 4 additions and 2 deletions

View File

@@ -90,7 +90,7 @@ def single_timeline(
work_queue.do(22, tenants, attach_broken)
env.pageserver.stop() # clears the failpoint as a side-effect
env.pageserver.stop(immediate=True) # clears the failpoint as a side-effect; immediate to avoid hitting neon_local's timeout
tenant_timelines = list(map(lambda tenant: (tenant, template_timeline), tenants))
log.info(f"python-side on-demand download the layer files into local tenant dir")
fixtures.pageserver.remote_storage.copy_all_remote_layer_files_to_local_tenant_dir(
@@ -100,7 +100,7 @@ def single_timeline(
log.info(f"wait for tenants to become active")
for tenant in tenants:
wait_until_tenant_active(ps_http, tenant)
wait_until_tenant_active(ps_http, tenant, iterations=ncopies, period=1)
# ensure all layers are resident for predictiable performance
for tenant in tenants:

View File

@@ -92,4 +92,6 @@ def test_getpage_throughput(
log.info(f"Results:\n{json.dumps(results, sort_keys=True, indent=2)}")
env.pageserver.stop(immediate=True) # with 20k tenants, we hit neon_local's shutdown timeout of 10 seconds
zenbenchmark.record_pagebench_results("get-page-latest-lsn", results, duration)