Instance Administration: Parameters, Memory & Workload

You created a HANA workload class capping statement memory for the SAP application, but it has no effect on those sessions. What step did you skip?

To cap statement memory and parallelism for a group of statements in HANA, do you use a workload class or a workload mapping — and what does each contribute?

A HANA query hits statement_memory_limit. Is it slowed down, queued, or killed — and why does that design choice matter?

A colleague says HANA log segments hold the database's table data. Correct them — what do log segments actually contain, and why does the distinction matter operationally?

During full-log recovery you move a log volume to a temp file system. Why is a symbolic link at the original path required for the database to start at all?

In M_LOG_SEGMENTS, how many segments per logging service can show state Writing at once, and what does that tell you when diagnosing?

Why does HANA normal log mode enable point-in-time recovery — what specifically does it guarantee that overwrite mode does not?

ALTER SYSTEM LOGGING OFF speeds up a big load by skipping redo writes. When is it justified, and what is the contraindication?

In M_LOG_SEGMENTS, which single state means a segment can be overwritten right now — Free, Closed, or BackedUp — and what separates it from the other two?

Connected to a tenant database, you run an ALTER CONFIGURATION on the SYSTEM layer intending to change a setting system-wide. What actually happens?

One tenant is consuming too much memory and starving the others. Do you lower global_allocation_limit or allocationlimit — and what's the difference?

A freshly installed HANA database runs in overwrite log mode and flips to normal by itself. What triggers the switch, and what's the underlying logic?

The log file system is full and the DB is frozen. What is the core recovery idea, and which tempting shortcut must you avoid?

A teammate wants to revert a customized HANA parameter to its default by SETting it to the default value. What's the correct way, and why is theirs subtly wrong?

A transaction is blocked on a DDL/metadata lock, not a row. Which view do you check — M_OBJECT_LOCKS or M_RECORD_LOCKS — and how do you keep them straight?

On a HANA system, a log segment shows state RetainedFree and is not being reused even though it's backed up. What is holding it, and where would you NEVER see this state?

To answer 'is HANA's footprint crowding the OS on this box?', do you read M_HOST_RESOURCE_UTILIZATION or a HANA memory-pool view — what does each level tell you?

Two HANA log segments are both un-backed-up and out of active use, one Closed and one Truncated. Which is closer to being reusable, and what's the single deciding difference?

After a full-log situation, you reclaim space with ALTER SYSTEM RECLAIM LOG. What is the one log-management action you must NEVER take, and why?

You ran ALTER SYSTEM ALTER CONFIGURATION ... SET without WITH RECONFIGURE and the new value isn't taking effect. Is the change lost, and what does WITH RECONFIGURE actually change?

global_allocation_limit shows its default value of 0. Does that mean 'no limit', and what memory does 0 actually grant HANA?

In HANA, the column store and the row store live in different memory types. Which sits in heap and which in shared memory — and which one dominates HANA's footprint?

What is the default value of log_mode in global.ini, and why might an admin wrongly believe it's 'overwrite'?

Users report new HANA requests being rejected with error 1038 under heavy load. What mechanism is responsible, and what is it doing?

A near-idle HANA database keeps creating new (sometimes large) log segments roughly every 15 minutes. Is something wrong, and what controls this?

Data-modifying statements hang AND you can't even open a new connection to HANA. What should be your first suspicion?

You suspect a redo-log problem (segments not freeing, log volume filling). Which view is the primary source for log state, and what columns do you read first?

Where does the HANA indexserver get its 1,024 MB default log segment size — and why is looking in indexserver.ini a dead end?

You just ran ALTER SYSTEM LOGGING ON after a load. Is the database recoverable yet, and what must you do before you can say yes?

You need to find who is connected to HANA right now and what statement each is running. Which view do you start from, and what does each row represent?

A parameter is set on both the SYSTEM and HOST layers in HANA with different values. Which value wins, and what's the full precedence order?

A production HANA system is running in overwrite log mode and crashes. How far back can you recover, and why should it never have been in this mode?

What is the default value of the HANA lock_wait_timeout parameter, and what happens when it's hit?

HANA cockpit / DBACOCKPIT shows log segments at 100% used even though they're nearly empty. Is your log full, and where's the real fill level?

Two systems both rely on the HANA redo log: crash recovery and system replication. What about the log's write timing makes both possible?