SOLR-18165 Complete dot-separated metric name migration

[janhoy]

https://issues.apache.org/jira/browse/SOLR-18165

PR #4223 converted the majority of Solr metric instrument names from underscore_delimited
to dot.separated OTel standard naming. Several files were missed or only partially converted.
This PR completes that work.

Files changed

AuditLoggerPlugin.java — all 7 metric names were still in old solr_auditlogger_* format;
converted to solr.auditlogger.*.

CaffeineCache.java — the default base metric name "solr_caffeine_cache" and the five
metric name suffixes (_lookups, _ops, _size, _ram_used, _warmup_time) were still using
underscores as namespace separators. Converted to "solr.caffeine_cache" base with
dot-prefixed suffixes (.lookups, .ops, .size, .ram_used, .warmup_time).

ZkContainer.java — all 7 metrics still in old solr_zk_* format; converted to solr.zk.*.

OtelMetrics.java (cross-dc-manager) — all 8 metric names were in old crossdc_consumer_*
format. The NAME_PREFIX constant is updated to "crossdc.consumer." and all suffix strings
are converted to dot-separated.

Metric name mapping

Old name	New name	Comments
solr_auditlogger_count	solr.auditlogger.count
solr_auditlogger_errors	solr.auditlogger.errors
solr_auditlogger_lost	solr.auditlogger.lost
solr_auditlogger_request_times	solr.auditlogger.request_times	Leaf kept as compound word
solr_auditlogger_queue	solr.auditlogger.queue
solr_auditlogger_queued_time	solr.auditlogger.queued_time	Leaf kept as compound word
solr_auditlogger_async_enabled	solr.auditlogger.async_enabled	Leaf kept as compound word
solr_caffeine_cache_lookups	solr.caffeine_cache.lookups	caffeine_cache is the cache type name
solr_caffeine_cache_ops	solr.caffeine_cache.ops
solr_caffeine_cache_size	solr.caffeine_cache.size
solr_caffeine_cache_ram_used	solr.caffeine_cache.ram_used	Leaf kept as compound word
solr_caffeine_cache_warmup_time	solr.caffeine_cache.warmup_time	Leaf kept as compound word
solr_zk_ops	solr.zk.ops
solr_zk_read	solr.zk.read
solr_zk_watches_fired	solr.zk.watches_fired	Leaf kept as compound word
solr_zk_written	solr.zk.written
solr_zk_cumulative_multi_ops_total	solr.zk.cumulative.multi_ops.total	Leaf kept as compound word;
solr_zk_child_fetches	solr.zk.child.fetches
solr_zk_cumulative_children_fetched	solr.zk.cumulative.children_fetched	Leaf kept as compound word
crossdc_consumer_input_msg_total	crossdc.consumer.input.msg.total	Consider _total?
crossdc_consumer_input_req_total	crossdc.consumer.input.req.total
crossdc_consumer_collapsed_total	crossdc.consumer.collapsed.total
crossdc_consumer_output_total	crossdc.consumer.output.total
crossdc_consumer_output_batch_size	crossdc.consumer.output.batch_size	Leaf kept as compound word
crossdc_consumer_output_backoff_time	crossdc.consumer.output.backoff_time	Leaf kept as compound word
crossdc_consumer_output_time	crossdc.consumer.output.time
crossdc_consumer_output_first_attempt_time	crossdc.consumer.output.first_attempt_time	Leaf kept as compound word
solr_zk_child_fetches	solr.zk.get_children.ops
solr_zk_cumulative_children_fetched		Discontinued

Questions:

• Do you agree with the name mappings above?
• The two existing solr_zk_child_fetches and solr_zk_cumulative_children_fetched both talk about the same thing, only cumulative is total? So a bit strange that they use child_fetches vs children_fetched. Should solr.zk.child.fetches instead be solr.zk.child_fetches to align with solr_zk_cumulative_children_fetched? Or should we do a breaking change here and align naming?
• The crossdc consumer metrics do not have a solr. prefix. The corresponsing producer metrics use solr.core.crossdc.producer.XXX. So ideally consumer would also use solr.core.crossdc.consumer.XXX. Or why the core part in there? Should we do a breaking change here, for one or for both?

[Copilot]

Pull request overview

Completes the remaining OpenTelemetry metric instrument renames from underscore-delimited names to dot-separated names as part of SOLR-18165.

Changes:

• Migrates AuditLoggerPlugin metric names to solr.auditlogger.*.
• Migrates CaffeineCache metric base name and suffixes to solr.caffeine_cache.*.
• Migrates ZooKeeper metrics in ZkContainer to solr.zk.*, and cross-dc consumer metrics to crossdc.consumer.*.
• Adds an unreleased changelog entry for this follow-up migration.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
solr/cross-dc-manager/src/java/org/apache/solr/crossdc/manager/consumer/OtelMetrics.java	Renames cross-dc consumer OTel instrument names to dot-separated format.
solr/core/src/java/org/apache/solr/security/AuditLoggerPlugin.java	Renames audit logger OTel instrument names to solr.auditlogger.*.
solr/core/src/java/org/apache/solr/search/CaffeineCache.java	Renames caffeine cache OTel instrument base name and suffixes to dot-separated format.
solr/core/src/java/org/apache/solr/core/ZkContainer.java	Renames ZooKeeper OTel instrument names to solr.zk.*.
changelog/unreleased/SOLR-18165-More-dot-separated-metric-names.yml	Adds changelog entry documenting the additional metric renames.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[janhoy]

@chan-dx I ping you here since I could not add you as reviewer...

[mlbiscoc]

Hey @janhoy sorry, just realized I missed this PR.

Do you agree with the name mappings above?

Mappings LGTM

The two existing solr_zk_child_fetches and solr_zk_cumulative_children_fetched both talk about the same thing, only cumulative is total? So a bit strange that they use child_fetches vs children_fetched. Should solr.zk.child.fetches instead be solr.zk.child_fetches to align with solr_zk_cumulative_children_fetched? Or should we do a breaking change here and align naming?

I took a deeper look at the collections and I don't think they talk about the same thing and I named them incorrectly. solr_zk_child_fetches fetched the number of getChildren calls:
childFetches.increment();

while solr_zk_cumulative_children_fetched got the number of actual child nodes across all of the operations.
cumulativeChildrenFetched.add(trace.getResponseChildrenCount());

WDYT of solr_zk_child_fetches -> solr.zk.get_children.ops and solr_zk_cumulative_children_fetched -> solr.zk.children.fetched? But then also would someone even care about the total number of solr.zk.children_fetched from a Solr perspective? I would probably just drop that metric now that I think of it but idk maybe someone thinks otherwise.

The crossdc consumer metrics do not have a solr. prefix. The corresponsing producer metrics use solr.core.crossdc.producer.XXX. So ideally consumer would also use solr.core.crossdc.consumer.XXX. Or why the core part in there? Should we do a breaking change here, for one or for both?

CC @sigram. We should add solr. prefix. Tbh I don't know enough about the crossdc module but to me seems like we don't need core there.

To answer your "breaking changes" question, yes I think we should just make the changes. The mailing list was already sent out about breaking changes on metrics so we should continue to go forward with it. I can see it being debatable if we did a drastic metric name overhaul across everything but a few here that are ZK metrics and the cross DC module should be fine.

[janhoy]

CC @sigram. We should add solr. prefix. Tbh I don't know enough about the crossdc module but to me seems like we don't need core there.

Proposal: Rename CrossDC metric names

Following @mlbiscoc's suggestion — add solr. prefix to consumer metrics and remove the spurious .core. segment from producer metrics. This makes both sides symmetric under solr.crossdc.*.

Consumer: crossdc.consumer. → solr.crossdc.consumer.

Old	New
crossdc.consumer.input.msg.total	solr.crossdc.consumer.input.msg.total
crossdc.consumer.input.req.total	solr.crossdc.consumer.input.req.total
crossdc.consumer.collapsed.total	solr.crossdc.consumer.collapsed.total
crossdc.consumer.output.total	solr.crossdc.consumer.output.total
crossdc.consumer.output.batch_size	solr.crossdc.consumer.output.batch_size
crossdc.consumer.output.backoff_time	solr.crossdc.consumer.output.backoff_time
crossdc.consumer.output.time	solr.crossdc.consumer.output.time
crossdc.consumer.output.first_attempt_time	solr.crossdc.consumer.output.first_attempt_time

Prometheus gains a solr_ prefix: crossdc_consumer_output_total → solr_crossdc_consumer_output_total

Producer: solr.core.crossdc.producer. → solr.crossdc.producer.

Old	New
solr.core.crossdc.producer.local.processed	solr.crossdc.producer.local.processed
solr.core.crossdc.producer.submitted	solr.crossdc.producer.submitted
solr.core.crossdc.producer.submitted.add	solr.crossdc.producer.submitted.add
solr.core.crossdc.producer.submitted.delete_by_id	solr.crossdc.producer.submitted.delete_by_id
solr.core.crossdc.producer.submitted.delete_by_query	solr.crossdc.producer.submitted.delete_by_query
solr.core.crossdc.producer.submitted.commit	solr.crossdc.producer.submitted.commit
solr.core.crossdc.producer.document_size	solr.crossdc.producer.document_size
solr.core.crossdc.producer.doc_too_large_errors	solr.crossdc.producer.doc_too_large_errors

Prometheus: solr_core_crossdc_producer_* → solr_crossdc_producer_*

[sigram]

@mlbiscoc yes, it makes sense to drop the "core" segment in the names.

[janhoy]

I pushed the crossdc changes. Will merge and backport this PR to 10x next week Tuesday if no objections.

[mlbiscoc]

LGTM! Thanks Jan for cleaning up some of my mess :)

[janhoy]

LGTM! Thanks Jan for cleaning up some of my mess :)

Not at all Matthew :) Looking forward to completing the new grafana dashboards...