Add OS-level performance tuning for Elasticsearch

[Oddly]

Description

The Elasticsearch role should optionally configure OS-level settings required for production deployments. Currently these must be handled externally (e.g. in the playbook or a separate OS role).

Settings to consider

Critical (bootstrap checks)

• vm.max_map_count = 262144 — required for mmapfs, ES refuses to start without it
• vm.swappiness = 1 — prevent JVM heap from being swapped to disk
• File descriptors (LimitNOFILE >= 65535) — already in RPM systemd unit
• Thread count (LimitNPROC >= 4096) — already in RPM systemd unit
• LimitMEMLOCK=infinity — already handled by the role when elasticsearch_memory_lock: true

Strongly recommended

• Disable Transparent Huge Pages (/sys/kernel/mm/transparent_hugepage/enabled = never)
• net.ipv4.tcp_retries2 = 5 — faster TCP timeout for cluster fault detection (~13s instead of ~15min)

Nice-to-have

• tuned profile throughput-performance — CPU governor, readahead, network buffers
• I/O scheduler: none for SSD, mq-deadline for HDD
• Network buffer tuning (somaxconn, rmem/wmem_max)

Suggested approach

Add an optional boolean (e.g. elasticsearch_os_tuning: false) that, when enabled, applies the critical and recommended sysctl settings via ansible.posix.sysctl and disables THP. This keeps the role self-contained for users who don't have a separate OS tuning role.

The tuned profile and I/O scheduler could be separate opt-in variables.

References

• https://www.elastic.co/guide/en/elasticsearch/reference/current/system-config.html
• https://www.elastic.co/guide/en/elasticsearch/reference/8.19/system-config-tcpretries.html