Reports 1-1 of 1 Clear search Modify search
DGS (General)
satoru.ikeda - 18:00 Tuesday 22 September 2026 (37511) Print this report
Recovery Status Following the Sep.21 Power Outage

Oshino-san, DanChen-san, Washimi-san, Ikeda

Related to K-Log#37506 

Below is a summary of the recovery work following the power outage on September 21.

DC Power Supplies

* All central DC power supplies for both the 24 V and 18 V systems had tripped.
* The DC power supplies at both end stations and on the second floor had not tripped and had recovered to a state in which power was being supplied.
* The front-panel power buttons on the V2 I/O chassis were OFF after power was restored. This is the expected behavior following a power outage.

FEPCs

* k1ix1, k1iy0, k1asc0, and k1test0 had restarted but stopped during the boot process because k1boot was unavailable and the required file systems could not be mounted.
* All other FEPCs were powered off.

The following recovery work was performed on the real-time systems:

1. After waiting for the air-conditioning system and k1boot to recover, we switched off the 18 V and 24 V power supplies. On the rack side, we also switched off the circuit breakers, chassis power supplies, and power strips.

2. We placed the control panels in operation mode.

3. For both the 24 V and 18 V systems, we performed the following sequence:

   * Switched on the power supply
   * Measured the voltage at the circuit breaker
   * Switched on the circuit breaker
   * Measured the voltage at the power strip
   * Switched on the power strip
   * Switched on the chassis power supply

4. After powering on all equipment in the central area, we followed the same procedure for the second floor of the central area and the first and second floors of both the X- and Y-end stations. We also powered on the workstations.

Some equipment, including the OMC PZT drivers, POS-related equipment, and several other devices, remains powered off.

The DC power supplies on the second floor of the central area and on the first and second floors of both end stations had remained on. However, all real-time PCs and I/O chassis had their power switches set to OFF, so we switched them on.

5. We restored the real-time computers in the central area.

Recovered:

* Without Dolphin:
  k1mcf0, k1ix1, k1iy1, k1ex1, k1ey1, k1ex0, k1ey0, k1test0, k1iy0

* With Dolphin:
  k1lsc0, k1asc0, k1als0, k1ioo, k1ioo1, k1imc0

Although some systems show drift in IRIGB_TIME, we confirmed that they have otherwise recovered and are operating normally.

Remaining systems:

* Without Dolphin:
  k1px1 — The electronics power supply has not yet been checked.

* With Dolphin:
  k1pr2, k1pr0, k1prm, k1bs, k1sr2, k1sr3, k1srm, k1omc0, k1omc1

The cards on k1pr2 are not being recognized. Due to time constraints, we will investigate the issue in detail tomorrow.
 

Non-image files attached to this report
Comments to this report:
shoichi.oshino - 18:10 Tuesday 22 September 2026 (37512) Print this report
Turned on the air conditioning in the morning before starting up k1boot. Room temperature had already risen to about 40°C at that point.

After confirming power was restored to the mine entrance, began DAQ recovery work. Started up DC, FW, TW, and NDS — DC was operating normally, but noticed that the MEDM displays for FW, TW, and NDS had some trouble.

Also noticed that the LED on DC's DAQ-out side NIC was off. Connected to the DAQ switch via serial console and found that a fault had occurred on the switch side.

Recovered by power-cycling the DAQ switch (cold boot — unplugging and replugging the power cable). FW, TW, and NDS then recovered automatically afterward.
satoru.ikeda - 15:53 Wednesday 23 September 2026 (37518) Print this report

Oshino-san, Ikeda,

We continued the recovery work for K-Log #37511 yesterday. All digital systems have now been restored.

### k1boot power investigation

k1boot is connected to the PDU and UPS. [The original sentence about the UPS battery is incomplete.]
The UPS is connected to k1boot, k1cam0, k1cam1, and k1det0. Among the systems other than k1boot, k1com had an uptime of about 8 days and k1det about 201 days.
The UPS and k1boot are connected by USB, and k1boot was shut down by apcupsd. The reason the time is shown as 02:00 still needs investigation. There were no logs from the previous momentary power outage on August 27.

Sep 21 02:08:21 k1boot apcupsd[4298]: Power failure.
Sep 21 02:08:27 k1boot apcupsd[4298]: Running on UPS batteries.
Sep 21 02:08:58 k1boot apcupsd[4298]: Reached run time limit on batteries.
Sep 21 02:08:58 k1boot apcupsd[4298]: Initiating system shutdown!
Sep 21 02:08:58 k1boot apcupsd[4298]: User logins prohibited
Sep 21 16:06:44 k1boot apcupsd[4303]: apcupsd 3.14.10 (13 September 2011) gentoo startup succeeded
Sep 21 16:06:44 k1boot apcupsd[4303]: NIS server startup succeeded

### DGS

We restored all remaining FrontEnd PCs (FEPCs). All digital systems are now running.

After all models were started, IRIGB_TIME settled at values outside the acceptable range. We restarted the IOP models, which resolved the issue.

* K1IOO, K1PR2, K1EY0: settled at 27
* K1SR3, K1SRM: settled at 43

We also started k1bcst and k1nds3, which had been missed during yesterday’s recovery work.

### HWP

After a power outage, the rotary encoders lose track of their positions, so their home positions must be found again. We found the home positions for PSL and IFO_REFL and initialized them to 0 deg.
IMC remains uninitialized at Ushiba-san’s request. It can still be operated by treating its position after the power outage—the current position—as 0 deg.
After logging in to the HWP control PC, we ran the following commands to find the home position:

$ cd /kagra/apps/agilis/
$ sudo ./agilis-p_controls_zero.py 1 1 1 RotateriSerial

Rotater iSerial values are listed here:

The command exits when the rotator reaches its home position. During normal operation, “After status” changes from 1E (moving) to 32 (finished).

After that, when the rotator is moved from MEDM, “total” indicates the angle moved from the home position above. It is 0 deg immediately after initialization. There is no need to write an initial position from MEDM or elsewhere.

We changed the initial value for the AS HWP, as addressed in K-Log #37328:

> 1. A temporary value of 150 was set to prevent the division-by-zero error.
>    [medm] sitemap → SYS → HWP OVERVIEW → count2degree

We wrote 150 to `K1:SYS-HWP_IFO_AS_CT_PER_DEG_{P,N}` using `caput`.
 

Non-image files attached to this comment
satoru.ikeda - 16:52 Wednesday 23 September 2026 (37519) Print this report

The IOP model defines two ADC cards for PR2, but there is only one physical card.

Since the extra card is unused, we’ll remove it from the IOP model at a later opportunity.

Non-image files attached to this comment
satoru.ikeda - 16:56 Wednesday 23 September 2026 (37520) Print this report

Restoring the Summary Page

Since k1sum0 had gone down, I turned on the power and confirmed that it was mounted via gwdet.

I confirmed that the summary page was restored.

Non-image files attached to this comment
Search Help
×

Warning

×