cliu
83af63ab53
Full manual diffs (C50, R30, C60, L20, C40, L35, R10, R20, J10, L30, L50, P30, G10, C10) and read-only checks on the shared R1.05 test server where the permission system allowed them. Skills and summaries moved to R1.05. Issues ISS-092..129 (gaps 104-107, 114-115, 130-131 unused); ISS-074 closed (manual and software agree), ISS-091 corrected by ISS-119. version-differences: unlisted manual changes and installed-only changes. r105-verification: V5-V9 and new not-tested rows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
3.6 KiB
task: Tuning checks (read-only) version: R1.05 (first written for R1.03) source: IM 36K10J10-01EN §5.5.1 (5th ed., Feb. 2026); practiced on the test system 2026-09-30 (R1.03) status: verified (read-only checks) on R1.03 (2026-09-30) and on the R1.05 test server (2026-10-08: setup-file limits, snapshot files, itmdbg code 2)
Tuning checks (read-only)
Checks whether CI Server needs tuning. Nothing here changes the system. Changing the limits means editing setup files and restarting CI Server: ask the user first.
Limits in setup files (%TLS_ROOT_PATH%\tls\sup\)
Read the active (non-!) lines:
| File | Keyword | Meaning | R1.03 and R1.05 test systems |
|---|---|---|---|
dur.sup |
dur_max_proc |
Max processes (default 250) | 250 |
dur.sup |
pool_kbytes |
DUR message pool | 250000 |
itm.sup |
pool_kbytes |
Item common memory pool | 650000 (itm.sup identical on R1.05) |
database_start.cmd (tls\com) |
-u / -f |
Max database users (default 2K) / files (default 1,200) | not checked |
Live values: automatic health snapshots
CI Server's system monitor (smdmon) writes debugger snapshots about every 10 minutes to
%TLS_LOG_PATH%\smd_<node>_<debugger>_*.log (e.g. smd_UNLICENCED_LOW_itmdbg_all.log).
Read the newest entries:
| File | What to check | Target (IM 36K10J10 §5.5.1) |
|---|---|---|
*_itmdbg_all.log |
common size / free size |
Free above 10 %, usage below 75 % |
*_smdsnp_process.log |
Per process: queue_size, queue_used, queue_high |
No requests waiting (queue_used 0); enlarge a queue that stays in use |
*_ismdbg_all.log |
Total number of lock collisions, users, files |
Collisions not rising above 250/s; entries below 80 % of max users |
*_smdsnp_process.log |
Number of processes | Well below dur_max_proc |
Test system (R1.03, 2026-09-30): common memory 665.6 MB, free 665.3 MB; queues empty; 3 lock collisions in total.
R1.05 (2026-10-08): same file naming with the node name (smd_<node>_LOW_<debugger>_all.log,
previous one kept as .old). New snapshots from the system monitor (smdmon_config.csv):
*_netstat_anob.log (open IP ports with their processes) and *_curl_isgsnap.log (a snapshot
from the web server at http://localhost:8080/isgsnap). The system monitor switches
umhlog logging on with size argument l40 (R1.03: l4); on the R1.05 test server
umhlog.old was about 40 MB, so the message log now rolls over at about 40 MB (R1.03 size
limit not observed).
The low-level snapshot script writes InstalledProducts.txt with PowerShell instead of wmic.
Live values from the interactive debuggers (read-only)
The debuggers ask for a code; pipe the answers in and end with -1 (exit). Use only the
read-only codes listed in the manual summaries:
| Command | Shows | Reference |
|---|---|---|
printf '2\n-1\n' | itmdbg |
item common size/free, number of items, integration granule, update counters | knowledge/manuals/IM36K10L36-01EN.md §4 |
printf '1\n-1\n' | eqpdbg 0 <EQP process> |
equipment manager: lines attached, scan tick, slave count, line load % | knowledge/manuals/IM36K10L40-01EN.md §3 |
Verified 2026-09-30 (bash, R1.03); itmdbg code 2 also on R1.05 (2026-10-08, piped from
PowerShell: common size 665.6 MB, free 665.1 MB). Other codes change the running system (delete items, unlock,
update scan status); do not use them without the user's approval.
Not yet practiced
Debug launchers (jwuipoi_debug, jhmiset_debug, …), other debuggers (opcdbg, ithdbg,
…), and per-process debug settings; see knowledge/manuals/IM36K10J10-01EN.md §5.5.2.