darold/pgcluu v5.0

Version 5.0

on GitHub

2025-10-19 - Version 5.0

This major release adds a lot of new reports and fix several issues
reported since last release.

New features:

• Add report about replication slots retrieved from pg_replication_slots.
• Add collect of pg_hba_file_rules
• Limit wait event reporting to those that represent more than 0.1 percent
  of the total of the wait event observed.
• Add collect of I/O statistics from pg_stat_io (PG16+)
• Add reports about:
  • I/O throughput per backend type
  • shared buffers writers and evictions
  • buffer hit ratio and ring buffers
  • I/O latency and size
  • WAL I/O
  • temporary relations I/O
  • asynchronous I/O (PG18+)
• Add collect more kernel tuning parameters, remove shmmax and shmall
  • vm.dirty_expire_centisecs and vm.dirty_writeback_centisecs, they were
    not matched by the sysctl filter, and all other vm.dirty_* parameters,
  • vm.min_free_kbytes, vm.watermark_scale_factor and kernel.numa_balancing,
  • net.core.somaxconn, net.ipv4.ip_local_port_range, net.core.rmem_max,
    net.core.wmem_max, fs.file-max and fs.nr_open,
  • /sys/kernel/mm/transparent_hugepage/shmem_enabled: PostgreSQL shared
    memory is anonymous shared memory, this is the setting that applies to
    the shared buffers, not 'enabled',
  • the frequency governor of all CPUs instead of cpu0 only, the cpuidle
    driver, the intel_pstate status and energy performance preference, and
    the number of NUMA nodes,
  • read_ahead_kb, nr_requests, max_sectors_kb and wbt_lat_usec of each
    block device.
• Collect and report the limits of the postmaster process. pgcluu_collectd now
  saves in a new POSTMASTER section of sysinfo.txt:
  • the pid of the postmaster, found as the parent of the checkpointer
    (its /proc entry is readable by any user) or read from postmaster.pid
    before PostgreSQL 10,
  • oom_score_adj, -1000 is recommended for the postmaster,
  • the soft and hard limits of open files, processes and locked memory,
  • the cgroup limits: memory.max, memory.high, memory.swap.max, cpu.max,
    cpu.weight, io.max, pids.max and the oom/oom_kill counters of
    memory.events with cgroup v2 (also in hybrid mode), memory and cpu
    quota limits with cgroup v1.
    pgcluu and pgcluu.cgi show them in the Kernel panel of the system
    information with sizes and CPU quotas formatted, and a tooltip for
    oom_score_adj and the process limits.
• Add kernel pressure stall information and memory events reports.
  pgcluu_collectd now appends at each run to kernel_stats.csv the cumulative
  counters of:
  • the pressure stall information (PSI) of the host (/proc/pressure, kernel
    4.20+) and of the cgroup v2 of the postmaster (cpu, memory and io
    pressure),
  • the memory events of the cgroup of the postmaster (high, max, oom and
    oom_kill from memory.events),
  • /proc/vmstat counters not reported by sar: direct reclaims (allocstall),
    memory compaction, transparent huge pages allocations and fallbacks,
    OOM kills, NUMA allocations and page cache refaults.
  • pgcluu and pgcluu.cgi have two new pages in the System menu:
    • Pressure stall information: percentage of time with some or all tasks
      stalled on CPU, memory or I/O for the host and the PostgreSQL cgroup,
    • Kernel memory events: allocation stalls and compaction, transparent huge
      pages, OOM kills, cgroup memory events, page cache refaults and NUMA
      allocations.

Here is the full list of fixes since previous release.

• Fix PostgreSQL version and uptime missing from sysinfo.txt
• Add PG17 compatibity: rebuild pg_stat_bgwriter. Thanks to Christophe
  Courtois for the patch.
• Fix postgresql iodelay report.
• Fix System memory utilization report, since sar reports kbavail in sar
  output.
• Fix sar await position detection with sysstat < v12.1.1.
  Thanks to Scott Dellaperuta for the report.
• Fix pgcluu reports in capture mode complaining about non existing end-*
  files. Thanks to Stephan Fercot for the report.
• Fix pgcluu reports in capture mode when the host is in streaming
  replication.
• Fix pidstat reports: values read by column name and summed over the processes
  The pidstat reports (PostgreSQL CPU, memory, page faults, I/O and context
  switches) were wrong with the output of the pidstat command run by
  pgcluu_collectd.
• Fix concurrent cache runs, cache files read during their rewrite and
  input directory normalization. pgcluu -C run by cron every five minutes
  could start while the previous run was still writing the cache files,
  and pgcluu.cgi could read them at the same time.
• The cluster-conflicts page of the CGI showed a graph for each database in
  addition to the cluster total, it now only shows the total. The graph of
  a database is shown by the database-conflicts action.
• Fix cluster totals of tablespaces, prepared transactions, buffercache and conflicts
• Fix double ALL global connections and database size for cluster on charts. Thanks to MirekWa for the report.
• Fix Apache activation path for custom installations. Thanks to Wu Shuwen for the patch.
• Fix documentation formatting for pgcluu options.
• Fix report CSS style to extend array length to not truncate system information.
• Fix lock graphs whose data points were not ordered by time
  The series of the lock types, lock modes and granted locks graphs were built
  by iterating over the timestamps in hash order, so the lines of the graphs
  were drawn back and forth in time.
• pgcluu.cgi: only load the cache files required by the requested report