Debugging the Storage Manager
[SSMDEBUG]

Collaboration diagram for Debugging the Storage Manager:

---

Detailed Description

The storage manager contains a few methods that are useful for debugging purposes. Some of these should be used for not other purpose, as they are not thread-safe, or might be very expensive.

Functions
static rc_t	ss_m::dump_xcts (ostream &o)
	Print transaction information to an output stream.
static rc_t	ss_m::checkpoint ()
	Take a checkpoint.
static rc_t	ss_m::force_buffers (bool invalidate=false)
	Force the buffer pool to flush its pages to disk.
static rc_t	ss_m::force_vol_hdr_buffers (const vid_t &vid)
	Force the buffer pool to flush the volume header page(s) to disk.
static rc_t	ss_m::force_store_buffers (const stid_t &stid, bool invalidate)
	Force the buffer pool to flush to disk all pages for the given store.

---

Function Documentation

static rc_t ss_m::dump_xcts

(

ostream &

o

)

[static, inherited]

Print transaction information to an output stream.

Parameters:

[in]

o

Stream to which to write the information.

This is for debugging only, and is not thread-safe.

static rc_t ss_m::checkpoint

(

)

[static, inherited]

Take a checkpoint.

Note:

For debugging only!

Force the storage manager to take a checkpoint. Checkpoints are fuzzy : they can be taken while most other storage manager activity is happening, even though they have to be serialized with respect to each other, and with respect to a few other activities.

This is thread-safe.

static rc_t ss_m::force_buffers

(

bool

invalidate = false

)

[static, inherited]

Force the buffer pool to flush its pages to disk.

Parameters:

[in]

invalidate

True means discard pages after flush.

Note:

For debugging only!

Attention:

Do not call force_buffers with anything pinned. You may cause latch-latch deadlocks, as this method has to scan the entire buffer pool and possibly EX-latch pages to prevent others from updating while it forces to disk. Since the page-order is essentially random, we cannot preclude latch-latch deadlocks with other threads.

static rc_t ss_m::force_vol_hdr_buffers

(

const vid_t &

vid

)

[static, inherited]

Force the buffer pool to flush the volume header page(s) to disk.

Parameters:

[in]

vid

ID of the volume of interest

Note:

For debugging only!

Attention:

Do not call force_vol_hdr_buffers with anything pinned. You could cause latch-latch deadlocks, as this method has to scan the entire buffer pool and possibly EX-latch some pages. Since the page-order is essentially random, we cannot preclude latch-latch deadlocks with other threads.

static rc_t ss_m::force_store_buffers	(	const stid_t &	stid,
		bool	invalidate
	)			[static, inherited]

Force the buffer pool to flush to disk all pages for the given store.

Parameters:

[in]	stid	Store whose pages are to be flushed.
[in]	invalidate	True means discard the pages after flushing.

Note:

For debugging only!

Attention:

Do not call force_store_buffers with anything pinned. You may cause latch-latch deadlocks, as this method has to scan the entire buffer pool and, if invalide==true, EX-latch pages to prevent others from updating while it forces to disk. Since the page-order is essentially random, we cannot preclude latch-latch deadlocks with other threads.

---