On August 24, Typhoon Yangliu passed through Guangzhou, and a large area of power failure occurred in Panyu District. A dental clinic called us that night: the power failure lasted for 12 minutes. After the call, the HIS system of the clinic (registration, medical records and charges were all in it) could not be opened. The server was Lenovo ThinkSystem ST250, and the SQL Server database consistency error was reported.
The person in charge of the clinic did not understand: "I bought UPS five years ago for more than RMB 10,000. The indicator light has always been green. How can there be a power failure?"
We on-site on the morning of August 25 to repair the database (described later) and focus on the UPS problem-because this is the root cause of this accident and a hidden danger that many companies have.
UPS green light will lie: it only means "mains normal"
The green light on the panel of the on-site Natai special C3KS (3kVA online) is always on. However, many users have the wrong understanding of the indicator light:
| indicator status | true meaning | common misconceptions |
|---|---|---|
| The mains green light is always on | The input mains is normal, UPS is charging the battery (charging circuit works, does not mean that the battery can store electricity) | UPS is OK. |
| Battery light/fault light on | Abnormal battery voltage or protection triggered | Most machines don't light this light until they're dead. |
| Backup time displayed on panel | Most models areestimateAccording to the nominal capacity of the new battery, it does not reflect the actual attenuation of the battery | "The panel says it can last 30 minutes, it can really last 30 minutes." |
In other words:A battery has been completely scrapped UPS, as long as plugged in the mains, the green light is still bright every day.The only reliable way to tell is to disconnect the mains for an actual discharge test, or to test the battery with an internal resistance meter.
Actual measurement: battery internal resistance exceeds 3 times, actually can only last 3 minutes
We measure section by section with internal resistance tester (C3KS built-in 6 12V9AH maintenance-free lead-acid batteries):
| battery number | float voltage | Internal resistance (mΩ) | decide |
|---|---|---|---|
| #1 | 13.5V | 48 | Exceeded standard (new battery approx. 15mΩ) |
| #2 | 13.4V | 62 | seriously exceeded |
| #3 | 12.9V | 105 | scrap |
| #4 | 13.3V | 71 | seriously exceeded |
| #5 | 13.5V | 55 | overproof |
| #6 | 13.1V | 88 | scrap |
Lead-acid batteries should be replaced if their internal resistance exceeds the factory value by 1.5-2 times. This batch exceeds the standard by 3-7 times on average. Actual discharge verification has been done: disconnect the mains, with server + switch (measured load 780W),UPS alarm shutdown after 2 minutes and 50 seconds. The nominal 3kVA with 780W load, the new battery should last 20-25 minutes.
The battery has been used for 5 years. Under the environment of Guangzhou server room (summer room temperature 32- 3 - 5 ℃, long-term high temperature), the life of lead-acid battery is discounted. The industry standard is 3-5 years mandatory replacement. This one is overdue and has never been tested.
Maintenance: Database repair + battery replacement
database portion: SQL Server reported page 823/824 error, which is a corrupt transaction page being written when the power is off. We used DBCC CHECKDB to locate 3 corrupt pages, and recovered to the state 15 minutes before the accident through the last full backup (the previous Sunday)+ transaction log roll-forward. The registration record of 15 minutes that night was lost (manual supplement). Total time spent 6 hours, charge RMB 2800.
By the way, he discovered that the clinic's automatic backup task failed three months ago (the backup target disk was full and no one cared). If the damage was more serious this time, the data that could be recovered would have to be returned to three months ago.Backup tasks must be equipped with failure warnings, otherwise "there are backups" is only psychological comfort.
battery portion: 6 12V9AH maintenance-free batteries (Sante original equipment), including replacement, discharge test, waste battery recycling, quoted RMB 1180. At the same time, it is recommended to install a network management card (SNMP card, RMB 680), configure mail alarm-mains power failure, battery abnormality, load overrun will send mail, this is the key accessories to change UPS from "dumb equipment" to "alarm equipment".
How to calculate UPS capacity: Don't buy small or big
A simple solution for SMEs:
- Count the load power to be protected: Server nameplate power × number (actual power consumption is about 60% of nameplate)+ switch/router + necessary display equipment. Example: 2 servers nameplate 550W (actual about 330W/set)+ switch 80W = 740W.
- Calculate UPS capacity: Load power <$power factor (online type generally 0.8-0.9)<$0.8 (leave 20% margin). 740W <$0.85 <$0.8 ≈ 1090VA, select 2kVA gear.
- Calculate the backup time: The battery pack capacity depends on the required battery life. 2kVA standard machine (built-in battery) with 740W can last about 12-18 minutes; to last 1 hour, you must connect the external battery rack.
- Key question: How long will it last?It's not the longer the better. It depends on the target--Safe shutdown onlyFifteen minutes is enough;To wait for service to switch to generator or mains to resumeIt takes 1-4 hours to connect the battery rack externally;To keep the business completely uninterruptedIt requires a combination of two-way mains + generator + long delay UPS.
Common mistakes: bought a 3kVA long delay UPS but only connected to the built-in small battery pack (long delay model built-in battery is often only enough for 5 minutes, must be connected to the external battery rack); or UPS when the voltage regulator, connected to the air conditioner printer such a large inductive load (start current shock will instantly defeat UPS).
On-line or backup?
| type | theory | switching period | apply | Price (1kVA) |
|---|---|---|---|---|
| Off-line (Offline) | Usually the mains is connected, and the battery is cut off when the power is off. | 4-10 millisecond | Home computers, single POS terminals | RMB 300–500 |
| Line Interactive | With voltage regulation, power cut battery | 2-6 millisecond | Small office, NAS | RMB 600–1000 |
| Online (Online/Double Transformation) | mains → dc → inverter, zero switching | 0 ms | Servers, server rooms, medical equipment | $1500-$3000 |
The server and NAS are all online: zero switching time means that the load is not perceived at the moment of power failure, and the online type comes with complete mains filtering and voltage regulation, which is especially important for Guangzhou's summer voltage fluctuation environment.
UPS quarterly self-check list (10 minutes)
- mission panel: Input/output voltage, load percentage, battery voltage is in normal range, whether there is alarm history.
- Listen to sounds, smell smells: Fan abnormal noise, capacitor drum leakage will have sour taste, battery compartment has peculiar smell immediately cut off power inspection.
- Make a real discharge.: Disconnect the mains (unplug or open the air), with the actual load to observe how long it can last. If it is lower than 60% of the nominal backup time, replace the battery.It's the only way not to be fooled by the green light.
- Record ambient temperature: Lead-acid battery life halved at every 10℃ above 25℃. The air conditioner in the server room broke down for three months, and the battery life may have been halved.
- Check battery date: There is usually a factory date on the body or battery. If it is more than 4 years old, it will enter the replacement budget. Don't wait for it to expire.
ACCPC provides UPS full range of services: battery detection (internal resistance test report, free), battery replacement (including waste battery environmental recycling), UPS host maintenance and rental, power supply protection scheme design and construction. 020-39029800.