Files
neon/test_runner/fixtures/metrics.py
Christian Schwarz 79401638df remove materialized page cache (#8105)
part of Epic https://github.com/neondatabase/neon/issues/7386

# Motivation

The materialized page cache adds complexity to the code base, which
increases the maintenance burden and risk for subtle and hard to
reproduce bugs such as #8050.

Further, the best hit rate that we currently achieve in production is ca
1% of materialized page cache lookups for
`task_kind=PageRequestHandler`. Other task kinds have hit rates <0.2%.

Last, caching page images in Pageserver rewards under-sized caches in
Computes because reading from Pageserver's materialized page cache over
the network is often sufficiently fast (low hundreds of microseconds).
Such Computes should upscale their local caches to fit their working
set, rather than repeatedly requesting the same page from Pageserver.

Some more discussion and context in internal thread
https://neondb.slack.com/archives/C033RQ5SPDH/p1718714037708459

# Changes

This PR removes the materialized page cache code & metrics.

The infrastructure for different key kinds in `PageCache` is left in
place, even though the "Immutable" key kind is the only remaining one.
This can be further simplified in a future commit.

Some tests started failing because their total runtime was dependent on
high materialized page cache hit rates. This test makes them
fixed-runtime or raises pytest timeouts:
* test_local_file_cache_unlink
* test_physical_replication
* test_pg_regress

# Performance

I focussed on ensuring that this PR will not result in a performance
regression in prod.

* **getpage** requests: our production metrics have shown the
materialized page cache to be irrelevant (low hit rate). Also,
Pageserver is the wrong place to cache page images, it should happen in
compute.
* **ingest** (`task_kind=WalReceiverConnectionHandler`): prod metrics
show 0 percent hit rate, so, removing will not be a regression.
* **get_lsn_by_timestamp**: important API for branch creation, used by
control pane. The clog pages that this code uses are not
materialize-page-cached because they're not 8k. No risk of introducing a
regression here.

We will watch the various nightly benchmarks closely for more results
before shipping to prod.
2024-06-20 11:56:14 +02:00

156 lines
5.8 KiB
Python

from collections import defaultdict
from typing import Dict, List, Optional, Tuple
from prometheus_client.parser import text_string_to_metric_families
from prometheus_client.samples import Sample
from fixtures.log_helper import log
class Metrics:
metrics: Dict[str, List[Sample]]
name: str
def __init__(self, name: str = ""):
self.metrics = defaultdict(list)
self.name = name
def query_all(self, name: str, filter: Optional[Dict[str, str]] = None) -> List[Sample]:
filter = filter or {}
res = []
for sample in self.metrics[name]:
try:
if all(sample.labels[k] == v for k, v in filter.items()):
res.append(sample)
except KeyError:
pass
return res
def query_one(self, name: str, filter: Optional[Dict[str, str]] = None) -> Sample:
res = self.query_all(name, filter or {})
assert len(res) == 1, f"expected single sample for {name} {filter}, found {res}"
return res[0]
class MetricsGetter:
"""
Mixin for types that implement a `get_metrics` function and would like associated
helpers for querying the metrics
"""
def get_metrics(self) -> Metrics:
raise NotImplementedError()
def get_metric_value(
self, name: str, filter: Optional[Dict[str, str]] = None
) -> Optional[float]:
metrics = self.get_metrics()
results = metrics.query_all(name, filter=filter)
if not results:
log.info(f'could not find metric "{name}"')
return None
assert len(results) == 1, f"metric {name} with given filters is not unique, got: {results}"
return results[0].value
def get_metrics_values(
self, names: list[str], filter: Optional[Dict[str, str]] = None, absence_ok=False
) -> Dict[str, float]:
"""
When fetching multiple named metrics, it is more efficient to use this
than to call `get_metric_value` repeatedly.
Throws RuntimeError if no metrics matching `names` are found, or if
not all of `names` are found: this method is intended for loading sets
of metrics whose existence is coupled.
If it's expected that there may be no results for some of the metrics,
specify `absence_ok=True`. The returned dict will then not contain values
for these metrics.
"""
metrics = self.get_metrics()
samples = []
for name in names:
samples.extend(metrics.query_all(name, filter=filter))
result = {}
for sample in samples:
if sample.name in result:
raise RuntimeError(f"Multiple values found for {sample.name}")
result[sample.name] = sample.value
if not absence_ok:
if len(result) != len(names):
log.info(f"Metrics found: {metrics.metrics}")
raise RuntimeError(f"could not find all metrics {' '.join(names)}")
return result
def parse_metrics(text: str, name: str = "") -> Metrics:
metrics = Metrics(name)
gen = text_string_to_metric_families(text)
for family in gen:
for sample in family.samples:
metrics.metrics[sample.name].append(sample)
return metrics
def histogram(prefix_without_trailing_underscore: str) -> List[str]:
assert not prefix_without_trailing_underscore.endswith("_")
return [f"{prefix_without_trailing_underscore}_{x}" for x in ["bucket", "count", "sum"]]
PAGESERVER_PER_TENANT_REMOTE_TIMELINE_CLIENT_METRICS: Tuple[str, ...] = (
"pageserver_remote_timeline_client_calls_started_total",
"pageserver_remote_timeline_client_calls_finished_total",
"pageserver_remote_physical_size",
"pageserver_remote_timeline_client_bytes_started_total",
"pageserver_remote_timeline_client_bytes_finished_total",
)
PAGESERVER_GLOBAL_METRICS: Tuple[str, ...] = (
"pageserver_storage_operations_seconds_global_count",
"pageserver_storage_operations_seconds_global_sum",
"pageserver_storage_operations_seconds_global_bucket",
"pageserver_unexpected_ondemand_downloads_count_total",
"libmetrics_launch_timestamp",
"libmetrics_build_info",
"libmetrics_tracing_event_count_total",
"pageserver_page_cache_read_hits_total",
"pageserver_page_cache_read_accesses_total",
"pageserver_page_cache_size_current_bytes",
"pageserver_page_cache_size_max_bytes",
"pageserver_getpage_reconstruct_seconds_bucket",
"pageserver_getpage_reconstruct_seconds_count",
"pageserver_getpage_reconstruct_seconds_sum",
*[f"pageserver_basebackup_query_seconds_{x}" for x in ["bucket", "count", "sum"]],
*histogram("pageserver_smgr_query_seconds_global"),
*histogram("pageserver_layers_visited_per_read_global"),
*histogram("pageserver_getpage_get_reconstruct_data_seconds"),
*histogram("pageserver_wait_lsn_seconds"),
*histogram("pageserver_remote_operation_seconds"),
*histogram("pageserver_io_operations_seconds"),
"pageserver_tenant_states_count",
)
PAGESERVER_PER_TENANT_METRICS: Tuple[str, ...] = (
"pageserver_current_logical_size",
"pageserver_resident_physical_size",
"pageserver_io_operations_bytes_total",
"pageserver_last_record_lsn",
"pageserver_standby_horizon",
"pageserver_smgr_query_seconds_bucket",
"pageserver_smgr_query_seconds_count",
"pageserver_smgr_query_seconds_sum",
"pageserver_storage_operations_seconds_count_total",
"pageserver_storage_operations_seconds_sum_total",
"pageserver_evictions_total",
"pageserver_evictions_with_low_residence_duration_total",
"pageserver_aux_file_estimated_size",
*PAGESERVER_PER_TENANT_REMOTE_TIMELINE_CLIENT_METRICS,
# "pageserver_directory_entries_count", -- only used if above a certain threshold
# "pageserver_broken_tenants_count" -- used only for broken
)