List of Run-Time Options
[Run-Time Options]

Collaboration diagram for List of Run-Time Options:

These are the run-time options for the storage manager.

-sm_bufpoolsize :

• type: number
• description: This is the size of the buffer pool in Kb. Must be large enough to hold at least 32 pages, so it depends on the configured page size.
• default: none
• required?: yes

-sm_hugetlbfs_path

• type: string (full absolute path name)
• description: Needed only if you configured --with-hugetlbfs.
• default: see Configuration Options
• required?: no

-sm_reformat_log

• type: Boolean
• description: If "yes", your log will be clobbered and the storage manager will start up with an entirely new log.
• default: no
• required?: no

-sm_logdir

• type: string (relative or absolutee path name)
• description: Location of the log files.
• default: none
• required?: yes

-sm_logbufsize

• type: number
• description: size of log buffer in KB. Must be greater than or equal to the larger of (4 times the page size, 64 Kb) and less than or equal to 128 times the page_size. This is the size of the log buffer in Kb.
• default: 128
• required?: no

-sm_logsize

• type: number
• description: greater than or equal to 8256 This is the maximum size of the log in Kb. It is a function of the log buffer size, and the default is the minimum allowable for the default sm_logbufsize.
• default: 128
• required?: yes

-sm_log_warn

• type: number between 0 and 100 (percentage)
• description: percentage of log that, when consumed by active transactions, triggers a callback warning of potential inability to roll back. Should be less than 50.
• default: 45
• required?: no

-sm_errlog

• type: string (relative or absolute path name OR - )
• description: Destination for error messages. If "-" is given, the destination is stderr.
• default: -
• required?: no

-sm_errlog_level

• type: string (one of none|emerg|fatal|internal|error|warning|info|debug)
• description: filter. Message of this priority or higher are issued to the error log; messages with lower priority are not issued. The priorities are listed from high to low. "none" means no logging will happen.
• default: error
• required?: no

-sm_locktablesize :

• type: number greater than or equal to 64
• description: size of lock manager's hash table will be a prime number near and greater than the given number.
• default: 64000 (yields a hash table with 65521 buckets)
• required?: no

-sm_lock_escalate_to_page_threshold

• type: number greater than or equal to 0
• description: after acquiring this many record locks on a page, the lock will be escalated to a page lock. A value of 0 disables escalation to a page lock.
• default: 5
• required?: no

-sm_lock_escalate_to_store_threshold

• type: number greater than or equal to 0
• description: after acquiring this many page locks on in a store, the lock will be escalated to a store lock. A value of 0 disables escalation to a store lock.
• default: 25
• required?: no

-sm_lock_escalate_to_volume_threshold

• type: number greater than or equal to 0
• description: after acquiring this many store locks on in a volume, the lock will be escalated to a volume lock. A value of 0 disables escalation to a volume lock.
• default: 0
• required?: no

-sm_cc_alg

• type: string (one of file | page | record | none)
• description: default locking granularity for file operations. This can be overridden on a per-transaction basis with ss_m::set_xct_lock_level().
• default: record
• required?: no

-sm_backgroundflush

• type: Boolean
• description: Enables background-flushing of volumes. Must be set to "yes" for sm_num_page_writers to have any effect.
• default: yes
• required?: no

-sm_num_page_writers

• type: number
• description: greater than or equal to 0; this is the number of background-flushing threads for each volume. If you have lots of threads, a huge buffer pool, and few volumes, you should increase this. If sm_backgroundflush is "no", this value is ignored.
• default: 2
• required?: no

-sm_prefetch

• type: Boolean
• description: Enables prefetching for scans.
• default: no
• required?: no

-sm_logging

• type: Boolean
• description: Allows you to turn off logging for a run of the storage manager. This is only for experimentation, to measure logging overhead in a limited way. Aborts, rollbacks and restart/recovery do not work without logging. Independent concurrent transactions using btrees might not work without logging (this is not well-tested). Each time you start the server, you had better start with a clean device or a device that resulted from a clean shutdown of the prior run.
• default: yes
• required?: no

-sm_lock_caching

• type: Boolean
• description: Enables caching of transaction locks in transaction. Can be turned off for experimentation. If no, the default is not to cache locks, but any transaction can turn on caching for itself by calling the ss_m method set_lock_cache_enable(bool enable).
• default: yes
• required?: no

See also:

Examples of Server Code

---