mirror of
https://github.com/neondatabase/neon.git
synced 2026-01-16 01:42:55 +00:00
270d3be507
part of https://github.com/neondatabase/neon/issues/5899 Problem ------- Before this PR, the time spent waiting on the throttle was charged towards the higher-level page_service metrics, i.e., `pageserver_smgr_query_seconds`. The metrics are the foundation of internal SLIs / SLOs. A throttled tenant would cause the SLI to degrade / SLO alerts to fire. Changes ------- - don't charge time spent in throttle towards the page_service metrics - record time spent in throttle in RequestContext and subtract it from the elapsed time - this works because the page_service path doesn't create child context, so, all the throttle time is recorded in the parent - it's quite brittle and will break if we ever decide to spawn child tasks that need child RequestContexts, which would have separate instances of the `micros_spent_throttled` counter. - however, let's punt that to a more general refactoring of RequestContext - add a test case that ensures that - throttling happens for getpage requests; this aspect of the test passed before this PR - throttling delays aren't charged towards the page_service metrics; this aspect of the test only passes with this PR - drive-by: make the throttle log message `info!`, it's an expected condition Performance ----------- I took the same measurements as in #6706 , no meaningful change in CPU overhead. Future Work ----------- This PR enables us to experiment with the throttle for select tenants without affecting the SLI metrics / triggering SLO alerts. Before declaring this feature done, we need more work to happen, specifically: - decide on whether we want to retain the flexibility of throttling any `Timeline::get` call, filtered by TaskKind - versus: separate throttles for each page_service endpoint, potentially with separate config options - the trouble here is that this decision implies changes to the TenantConfig, so, if we start using the current config style now, then decide to switch to a different config, it'll be a breaking change Nice-to-haves but probably not worth the time right now: - Equivalent tests to ensure the throttle applies to all other page_service handlers.
119 lines
4.1 KiB
Python
119 lines
4.1 KiB
Python
import json
|
|
import uuid
|
|
|
|
from anyio import Path
|
|
from fixtures.log_helper import log
|
|
from fixtures.neon_fixtures import NeonEnvBuilder, PgBin
|
|
from fixtures.pg_version import PgVersion
|
|
from fixtures.types import TenantId, TimelineId
|
|
from fixtures.utils import wait_until
|
|
|
|
|
|
def test_pageserver_getpage_throttle(neon_env_builder: NeonEnvBuilder, pg_bin: PgBin):
|
|
env = neon_env_builder.init_start()
|
|
|
|
env.pageserver.tenant_detach(env.initial_tenant)
|
|
|
|
env.pageserver.allowed_errors.append(
|
|
# https://github.com/neondatabase/neon/issues/6925
|
|
r".*query handler for.*pagestream.*failed: unexpected message: CopyFail during COPY.*"
|
|
)
|
|
|
|
tenant_id = TenantId.generate()
|
|
timeline_id = TimelineId.generate()
|
|
|
|
rate_limit_rps = 100
|
|
compaction_period = 5
|
|
env.pageserver.tenant_create(
|
|
tenant_id,
|
|
conf={
|
|
"compaction_period": f"{compaction_period}s",
|
|
"timeline_get_throttle": {
|
|
"task_kinds": ["PageRequestHandler"],
|
|
"initial": 0,
|
|
"refill_interval": "100ms",
|
|
"refill_amount": int(rate_limit_rps / 10),
|
|
"max": int(rate_limit_rps / 10),
|
|
"fair": True,
|
|
},
|
|
},
|
|
)
|
|
|
|
ps_http = env.pageserver.http_client()
|
|
|
|
ps_http.timeline_create(PgVersion.V16, tenant_id, timeline_id)
|
|
|
|
def run_pagebench_at_max_speed_and_get_total_requests_completed(duration_secs: int):
|
|
cmd = [
|
|
str(env.neon_binpath / "pagebench"),
|
|
"get-page-latest-lsn",
|
|
"--mgmt-api-endpoint",
|
|
ps_http.base_url,
|
|
"--page-service-connstring",
|
|
env.pageserver.connstr(password=None),
|
|
"--runtime",
|
|
f"{duration_secs}s",
|
|
f"{tenant_id}/{timeline_id}",
|
|
]
|
|
|
|
basepath = pg_bin.run_capture(cmd, with_command_header=False)
|
|
results_path = Path(basepath + ".stdout")
|
|
log.info(f"Benchmark results at: {results_path}")
|
|
|
|
with open(results_path, "r") as f:
|
|
results = json.load(f)
|
|
log.info(f"Results:\n{json.dumps(results, sort_keys=True, indent=2)}")
|
|
return int(results["total"]["request_count"])
|
|
|
|
log.info("warmup / make sure metrics are present")
|
|
run_pagebench_at_max_speed_and_get_total_requests_completed(2)
|
|
metrics_query = {
|
|
"tenant_id": str(tenant_id),
|
|
"timeline_id": str(timeline_id),
|
|
"smgr_query_type": "get_page_at_lsn",
|
|
}
|
|
metric_name = "pageserver_smgr_query_seconds_sum"
|
|
smgr_query_seconds_pre = ps_http.get_metric_value(metric_name, metrics_query)
|
|
assert smgr_query_seconds_pre is not None
|
|
|
|
marker = uuid.uuid4().hex
|
|
ps_http.post_tracing_event("info", marker)
|
|
_, marker_offset = wait_until(
|
|
10, 0.5, lambda: env.pageserver.assert_log_contains(marker, offset=None)
|
|
)
|
|
|
|
log.info("run pagebench")
|
|
duration_secs = 10
|
|
actual_ncompleted = run_pagebench_at_max_speed_and_get_total_requests_completed(duration_secs)
|
|
|
|
log.info("validate the client is capped at the configured rps limit")
|
|
expect_ncompleted = duration_secs * rate_limit_rps
|
|
delta_abs = abs(expect_ncompleted - actual_ncompleted)
|
|
threshold = 0.05 * expect_ncompleted
|
|
assert (
|
|
threshold / rate_limit_rps < 0.1 * duration_secs
|
|
), "test self-test: unrealistic expecations regarding precision in this test"
|
|
assert (
|
|
delta_abs < 0.05 * expect_ncompleted
|
|
), "the throttling deviates more than 5percent from the expectation"
|
|
|
|
log.info("validate that we logged the throttling")
|
|
|
|
wait_until(
|
|
10,
|
|
compaction_period / 10,
|
|
lambda: env.pageserver.assert_log_contains(
|
|
f".*{tenant_id}.*shard was throttled in the last n_seconds.*",
|
|
offset=marker_offset,
|
|
),
|
|
)
|
|
|
|
log.info("validate that the metric doesn't include throttle wait time")
|
|
smgr_query_seconds_post = ps_http.get_metric_value(metric_name, metrics_query)
|
|
assert smgr_query_seconds_post is not None
|
|
actual_smgr_query_seconds = smgr_query_seconds_post - smgr_query_seconds_pre
|
|
|
|
assert (
|
|
duration_secs >= 10 * actual_smgr_query_seconds
|
|
), "smgr metrics should not include throttle wait time"
|