After WiFi's online, it is a long-term operational phase and many of the problems are revealed: log disks are all over the place, text message switch balances are running out of people's minds, some AP solid upgrade has resulted in an incompatible authentication method, backup servers have not been restored since last switch……Without systematic monitoring and inspection mechanisms, these problems often await a large number of complaints from users.
What should be the level of surveillance system to cover?
The monitoring of certification systems is not a single dimension and needs to be covered simultaneously from the infrastructure, service levels and operational layers.
Infrastructure layer monitoring: server CPU, memory, disk usage, network interface traffic, database service status, backup tasks. These are standard controls for basic IT carriers and most of the control tools can be covered.——The certification log will be written continuously, either the system is failed if the disk is full (the compliance log is lost) or the authentication service collapses. Set the disk usage rate of 70/85/95/III to alert sufficient time before it reaches full load.
Service level monitoring: Status of the certification service process, RADIUS service port response, Portal service HTTP response, database connection pool usage, external dependence (SMS gateways, I.D. interfaces). This layer is required to be at the process level, not just machine survival as normal——The machine is alive but the authentication process is dead, and it's also malfunctioning.
Level of operations monitoring: authentication success rate, certification response time (P50/P95/P99), online user trends, and text message success rates. The level of business level indicators directly reflects actual experience of users and is the first window to discover anomalies. The certification success rate suddenly dropped from 99% to 90%, possibly with external dependence; the certification response time P99 suddenly rose, possibly with a slow search or increased pressure on databases.
The principles of the warning rules
There are two common areas of error in the warning rules: too many warnings (every small matter is reported, and the carrier begins to ignore them) and too few warnings (the real problem is not covered).
P0 (Running): The certification service is not available, the RADIUS port does not respond, and the database server is blocked.
P1 (in-hour processing): The certification success rate is reduced by more than 5%, the disk usage rate is over 85%, SMS switch balances are below threshold and authentication P99 delays exceed five seconds.
P2 (daily concern): Log writing rates are abnormal (which may indicate that the log has been compressed/archived), backup tasks fail, and online users differ by more than 20% from the same time in history. Such anomalies do not affect current operations but need to be checked on the following working day.
SMS Balance Management
The loss of text message balances leads to a failure of authentication, which is a laughter that occurs in fact.WiFi web certification systemAutomatic monitoring of text message balances is required: alerts are made when below X and then when below Y, to ensure that the renewal is known and processed before it is fully used.
If you send 1,000 text messages per day, it is appropriate to have 3,000 alert thresholds (about three days of consumption).
The SMS balance is monitored by a multi-server backup mechanism. When the main service provider sends messages, it automatically switch to a back-up service provider, which effectively reduces the risk of interruptions in SMS services.
Daily inspection lists
Automatic monitoring covers known anomalies, but some problems are difficult to express with the control indicators and require regular manual inspections."Daily quick patrols + weekly in-depth inspections"。
Daily fast tour (5-10 minutes): check yesterday's certification success and authentication trends, whether there are any abnormal fluctuations; view disk usage; check text message balances; confirm whether the backup task is successfully performed. These items can be made into a dashboard that looks after work every day.
Weekly in-depth inspection (30-60 minutes): complete validation process with test account numbers to verify that authentication methods are normal; check the alarm records and processing of this week ' s proceedings; check whether log writing is continuous (if there is a time gap); confirm that the main preset switch mechanism is not abnormal; see system software version and whether there are security patches to apply.
Change management and configuration baseline
Many of the failures in the authentication system come from changes.——The AAP solids have been upgraded in a way that is not compatible, the database parameters are adjusted to lower performance, and the Portal page has been re-written to introduce compatibility issues. No change management of the wiring team has been introduced, which makes it particularly painful to list with problems because it does not know what has recently changed.
The core of change management is that all changes to the configurations are recorded (what, when and by whom); pre-allocation backup; core functionality validated normal after a change; and problems quickly roll back. This does not require complex tools, but a shared change log form is sufficient, but discipline is required, each change has to be filled out and memory alone.
Configuration baselines are those that secure the critical configuration status of the system when it is properly operational and regularly compare actual configurations with baselines, and detect drifts for timely correction. For multiple door lock-ups, configuration baselines help ensure consistency in the configuration of the shops and avoid IT personnel at a particular store"I changed one parameter myself."However, Headquarters was not informed.
Surveillance and transport regulations are not designed to be seen, but to allow the system to function in a stable and long-term manner. The time spent on building this set of things is far more expensive than the time saved by late breakdowns and effort to process client complaints.