timestamp_utc: 2026-06-03T07:32:16Z
cluster: dyec0602bcl
headnode: i-0c36c39770c533c8e
compute_host: i192bigmem-dy-all-1

slurm:
  running:
    - job: 48
      rule: run_bclconvert_tile_shard
      shard: L003/0005_tiles0525-0654
      elapsed: 5:46
      host: i192bigmem-dy-all-1
      cpus: 192
      mem: 1200000M
  pending:
    - job: 49
      reason: Resources
    - job: 50
      reason: Priority
    - job: 51
      reason: Priority
    - job: 52
      reason: Priority
    - job: 53
      reason: Priority

outputs:
  fastq_count: 0
  done_count: 0
  report_count: 0

fsx:
  mount: /fsx
  source: 10.0.1.213@tcp:/xtlzxb4v
  type: lustre
  options: rw,noatime,checksum,flock,user_xattr,lruresize,lazystatfs,nouser_fid2path,verbose,encrypt
  capacity: 6.6T
  used: 218G
  available: 6.4T
  use_percent: 4

dev_shm:
  capacity: 1.2T
  used: 172K
  available: 1.2T
  use_percent: 1
  dayoa_bclconvert_size: 172K
  dayoa_bclconvert_file_count: 5

bcl_convert_process:
  pid: 16009
  stat: Il
  elapsed: 5:39
  cpu_percent: 0.0
  rss_kb: 400180
  threads: 66
  wchan: ldlm_completion_ast
  read_bytes: 52826112
  write_bytes: 0
  cancelled_write_bytes: 0
  tile_regex_shape: exact_plus_joined_tile_terms
  tile_regex_note: active shard command carries a long exact OR expression for tiles s_3_2235 through s_3_2366 with gaps.

interpretation:
  - The compute node is not CPU, memory, local disk, or /dev/shm constrained.
  - The BCL Convert process has opened L003 CBCL files but is not producing /dev/shm output.
  - The observed state is consistent with a Lustre client lock/metadata wait or BCL Convert/Lustre interaction, not raw FSx bandwidth saturation.
  - Moving the path from /fsx/run_dir_mounts to /run_dir_mounts would not improve performance if it is backed by the same Lustre filesystem.
