1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-21 13:34:40 +03:00
lvm2/lib/metadata/replicator_manip.c

687 lines
17 KiB
C
Raw Normal View History

/*
* Copyright (C) 2009-2010 Red Hat, Inc. All rights reserved.
*
* This file is part of LVM2.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU Lesser General Public License v.2.1.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "lib.h"
#include "locking.h"
#include "metadata.h"
#include "segtype.h"
2010-05-21 17:34:09 +04:00
#include "toolcontext.h"
lvmetad: two phase vg_update Previously, a command sent lvmetad new VG metadata in vg_commit(). In vg_commit(), devices are suspended, so any memory allocation done by the command while sending to lvmetad, or by lvmetad while updating its cache could deadlock if memory reclaim was triggered. Now lvmetad is updated in unlock_vg(), after devices are resumed. The new method for updating VG metadata in lvmetad is in two phases: 1. In vg_write(), before devices are suspended, the command sends lvmetad a short message ("set_vg_info") telling it what the new VG seqno will be. lvmetad sees that the seqno is newer than the seqno of its cached VG, so it sets the INVALID flag for the cached VG. If sending the message to lvmetad fails, the command fails before the metadata is committed and the change is not made. If sending the message succeeds, vg_commit() is called. 2. In unlock_vg(), after devices are resumed, the command sends lvmetad the standard vg_update message with the new metadata. lvmetad sees that the seqno in the new metadata matches the seqno it saved from set_vg_info, and knows it has the latest copy, so it clears the INVALID flag for the cached VG. If a command fails between 1 and 2 (after committing the VG on disk, but before sending lvmetad the new metadata), the cached VG retains the INVALID flag in lvmetad. A subsequent command will read the cached VG from lvmetad, see the INVALID flag, ignore the cached copy, read the VG from disk instead, update the lvmetad copy with the latest copy from disk, (this clears the INVALID flag in lvmetad), and use the correct VG metadata for the command. (This INVALID mechanism already existed for use by lvmlockd.)
2016-06-08 22:42:03 +03:00
#include "lvmetad.h"
/* Add lv as replicator_dev device */
int replicator_dev_add_rimage(struct replicator_device *rdev,
struct logical_volume *lv)
{
if (!lv || !rdev)
return_0;
if (lv_is_rimage(lv)) {
log_error("Logical volume %s is already part of other "
"replicator.", lv->name);
return 0;
}
if (rdev->lv) {
log_error("Logical volume %s can not be attached to an "
"already defined replicator device", lv->name);
return 0;
}
lv_set_hidden(lv);
lv->rdevice = rdev;
rdev->lv = lv;
return add_seg_to_segs_using_this_lv(lv, rdev->replicator_dev);
}
/* Remove lv from replicator_dev device */
struct logical_volume *replicator_dev_remove_rimage(struct replicator_device *rdev)
{
struct logical_volume *lv;
if (!rdev || !rdev->lv)
return_NULL;
lv = rdev->lv;
if (!remove_seg_from_segs_using_this_lv(lv, rdev->replicator_dev))
return_NULL;
/* FIXME: - check for site references */
rdev->lv = NULL;
lv->rdevice = NULL;
lv_set_visible(lv);
return lv;
}
int replicator_dev_add_slog(struct replicator_device *rdev,
struct logical_volume *slog)
{
if (!slog || !rdev)
return_0;
if (rdev->slog) {
log_error("Replicator device in site %s already has sync log.",
rdev->rsite->name);
return 0;
}
if (slog->rdevice) {
log_error("Sync log %s is already used by replicator %s.",
slog->name, slog->rdevice->rsite->replicator->name);
return 0;
}
lv_set_hidden(slog);
slog->rdevice = rdev;
rdev->slog = slog;
return add_seg_to_segs_using_this_lv(slog, rdev->replicator_dev);
}
struct logical_volume *replicator_dev_remove_slog(struct replicator_device *rdev)
{
struct logical_volume *lv;
if (!rdev)
return_NULL;
lv = rdev->slog;
if (!lv) {
log_error("Replicator device in site %s does not have sync log.",
rdev->rsite->name);
return NULL;
}
if (!remove_seg_from_segs_using_this_lv(lv, rdev->replicator_dev))
return_NULL;
rdev->slog = NULL;
lv->rdevice = NULL;
lv_set_visible(lv);
return lv;
}
int replicator_add_replicator_dev(struct logical_volume *replicator_lv,
struct lv_segment *replicator_dev_seg)
{
if (!replicator_lv)
return_0;
if (!(replicator_lv->status & REPLICATOR)) {
dm_list_init(&replicator_lv->rsites);
lv_set_hidden(replicator_lv);
replicator_lv->status |= REPLICATOR;
}
if (!replicator_dev_seg)
return 1;
if (replicator_dev_seg->replicator) {
log_error("Replicator device %s is already part of replicator.",
replicator_dev_seg->lv->name);
return 0;
}
replicator_dev_seg->replicator = replicator_lv;
return add_seg_to_segs_using_this_lv(replicator_lv, replicator_dev_seg);
}
/**
* Returns rimage ?? lv upon succeful detach of device
* entire LV entry should be removed by this crootall ??
*/
struct logical_volume *replicator_remove_replicator_dev(struct lv_segment *replicator_dev_seg)
{
struct logical_volume *lv = NULL;
log_error("FIXME: not implemented.");
#if 0
/* FIXME: - this is going to be complex.... */
if (!replicator_dev_seg)
return_NULL;
/* if slog or rimage - exit */
if (!remove_seg_from_segs_using_this_lv(lv, replicator_seg))
return_NULL;
replicator_seg->rlog_lv = NULL;
lv->status &= ~REPLICATOR_LOG;
lv_set_visible(lv);
#endif
return lv;
}
int replicator_add_rlog(struct lv_segment *replicator_seg,
struct logical_volume *rlog_lv)
{
if (!rlog_lv)
return_0;
if (rlog_lv->status & REPLICATOR_LOG) {
log_error("Rlog device %s is already used.", rlog_lv->name);
return 0;
}
lv_set_hidden(rlog_lv);
rlog_lv->status |= REPLICATOR_LOG;
replicator_seg->rlog_lv = rlog_lv;
return add_seg_to_segs_using_this_lv(rlog_lv, replicator_seg);
}
struct logical_volume *replicator_remove_rlog(struct lv_segment *replicator_seg)
{
struct logical_volume *lv;
if (!replicator_seg)
return_0;
if (!(lv = replicator_seg->rlog_lv)) {
log_error("Replog segment %s does not have rlog.",
replicator_seg->lv->name);
return NULL;
}
if (!remove_seg_from_segs_using_this_lv(lv, replicator_seg))
return_NULL;
replicator_seg->rlog_lv = NULL;
lv->status &= ~REPLICATOR_LOG;
lv_set_visible(lv);
return lv;
}
#if 0
/*
* Create new LV to pretend the original LV
* this target will have a 'replicator' segment
*/
int lv_add_replicator(struct logical_volume *origin, const char *rep_suffix)
{
struct logical_volume *rep_lv;
char *name;
size_t slen;
if (!(name = strstr(origin->name, rep_suffix))) {
log_error("Failed to find replicator suffix %s in LV name %s",
rep_suffix, origin->name);
return 0;
}
slen = (size_t)(name - origin->name);
name = alloca(slen + 1);
memcpy(name, origin->name, slen);
name[slen] = 0;
if ((rep_lv = find_lv(origin->vg, name))) {
rep_lv->status |= VIRTUAL;
return 1;
}
if (!(rep_lv = lv_create_empty(name, &origin->lvid,
LVM_READ | LVM_WRITE | VISIBLE_LV,
ALLOC_INHERIT, origin->vg)))
return_0;
if (!lv_add_virtual_segment(rep_lv, 0, origin->le_count,
get_segtype_from_string(origin->vg->cmd,
"error")))
return_0;
rep_lv->status |= VIRTUAL;
return 1;
}
int lv_remove_replicator(struct logical_volume *lv)
{
return 1;
}
#endif
/*
* Check all replicator structures:
* only non-clustered VG for Replicator
* only one segment in replicator LV
* site has correct combination of operation_mode parameters
* site and related devices have correct index numbers
* duplicate site names, site indexes, device names, device indexes
*/
int check_replicator_segment(const struct lv_segment *rseg)
{
struct replicator_site *rsite, *rsiteb;
struct replicator_device *rdev, *rdevb;
struct logical_volume *lv = rseg->lv;
int r = 1;
if (vg_is_clustered(lv->vg)) {
log_error("Volume Group %s of replicator %s is clustered",
lv->vg->name, lv->name);
return 0;
}
if (dm_list_size(&lv->segments) != 1) {
log_error("Replicator %s segment size %d != 1",
lv->name, dm_list_size(&lv->segments));
return 0;
}
dm_list_iterate_items(rsite, &lv->rsites) {
if (rsite->op_mode == DM_REPLICATOR_SYNC) {
if (rsite->fall_behind_timeout) {
log_error("Defined fall_behind_timeout="
"%d for sync replicator %s/%s.",
rsite->fall_behind_timeout, lv->name,
rsite->name);
r = 0;
}
if (rsite->fall_behind_ios) {
log_error("Defined fall_behind_ios="
"%d for sync replicator %s/%s.",
rsite->fall_behind_ios, lv->name, rsite->name);
r = 0;
}
if (rsite->fall_behind_data) {
log_error("Defined fall_behind_data="
FMTu64 " for sync replicator %s/%s.",
rsite->fall_behind_data, lv->name, rsite->name);
r = 0;
}
} else {
if (rsite->fall_behind_timeout && rsite->fall_behind_ios) {
log_error("Defined fall_behind_timeout and"
" fall_behind_ios for async replicator %s/%s.",
lv->name, rsite->name);
r = 0;
}
if (rsite->fall_behind_timeout && rsite->fall_behind_data) {
log_error("Defined fall_behind_timeout and"
" fall_behind_data for async replicator %s/%s.",
lv->name, rsite->name);
r = 0;
}
if (rsite->fall_behind_ios && rsite->fall_behind_data) {
log_error("Defined fall_behind_ios and"
" fall_behind_data for async replicator %s/%s.",
lv->name, rsite->name);
r = 0;
}
if (!rsite->fall_behind_ios &&
!rsite->fall_behind_data &&
!rsite->fall_behind_timeout) {
log_error("fall_behind_timeout,"
" fall_behind_ios and fall_behind_data are"
" undefined for async replicator %s/%s.",
lv->name, rsite->name);
r = 0;
}
}
dm_list_iterate_items(rsiteb, &lv->rsites) {
if (rsite == rsiteb)
break;
if (strcasecmp(rsite->name, rsiteb->name) == 0) {
log_error("Duplicate site name "
"%s detected for replicator %s.",
rsite->name, lv->name);
r = 0;
}
if ((rsite->vg_name && rsiteb->vg_name &&
strcasecmp(rsite->vg_name, rsiteb->vg_name) == 0) ||
(!rsite->vg_name && !rsiteb->vg_name)) {
log_error("Duplicate VG name "
"%s detected for replicator %s.",
(rsite->vg_name) ? rsite->vg_name : "<local>",
lv->name);
r = 0;
}
if (rsite->site_index == rsiteb->site_index) {
log_error("Duplicate site index %d detected "
"for replicator site %s/%s.",
rsite->site_index, lv->name,
rsite->name);
r = 0;
}
if (rsite->site_index > rseg->rsite_index_highest) {
log_error("Site index %d > %d (too high) "
"for replicator site %s/%s.",
rsite->site_index,
rseg->rsite_index_highest,
lv->name, rsite->name);
r = 0;
}
}
dm_list_iterate_items(rdev, &rsite->rdevices) {
dm_list_iterate_items(rdevb, &rsite->rdevices) {
if (rdev == rdevb)
break;
if (rdev->slog && (rdev->slog == rdevb->slog)) {
log_error("Duplicate sync log %s "
"detected for replicator %s.",
rdev->slog->name, lv->name);
r = 0;
}
if (strcasecmp(rdev->name, rdevb->name) == 0) {
log_error("Duplicate device name %s "
"detected for replicator %s.",
rdev->name, lv->name);
r = 0;
}
if (rdev->device_index == rdevb->device_index) {
log_error("Duplicate device index %"
PRId64 " detected for "
"replicator site %s/%s.",
rdev->device_index,
lv->name, rsite->name);
r = 0;
}
if (rdev->device_index > rseg->rdevice_index_highest) {
log_error("Device index %" PRIu64
" > %" PRIu64 " (too high) "
"for replicator site %s/%s.",
rdev->device_index,
rseg->rdevice_index_highest,
lv->name, rsite->name);
r = 0;
}
}
}
}
return r;
}
/**
* Is this segment part of active replicator
*/
int lv_is_active_replicator_dev(const struct logical_volume *lv)
{
return ((lv->status & REPLICATOR) &&
lv->rdevice &&
lv->rdevice->rsite &&
lv->rdevice->rsite->state == REPLICATOR_STATE_ACTIVE);
}
/**
* Is this LV replicator control device
*/
int lv_is_replicator(const struct logical_volume *lv)
{
return ((lv->status & REPLICATOR) &&
!dm_list_empty(&lv->segments) &&
seg_is_replicator(first_seg(lv)));
}
/**
* Is this LV replicator device
*/
int lv_is_replicator_dev(const struct logical_volume *lv)
{
return ((lv->status & REPLICATOR) &&
!dm_list_empty(&lv->segments) &&
seg_is_replicator_dev(first_seg(lv)));
}
/**
* Is this LV replicated origin lv
*/
int lv_is_rimage(const struct logical_volume *lv)
{
return (lv->rdevice && lv->rdevice->lv == lv);
}
/**
* Is this LV sync log
*/
int lv_is_slog(const struct logical_volume *lv)
{
return (lv->rdevice && lv->rdevice->slog == lv);
}
/**
* Returns first replicator-dev in site in case the LV is replicator-dev,
* NULL otherwise
*/
struct logical_volume *first_replicator_dev(const struct logical_volume *lv)
{
struct replicator_device *rdev;
struct replicator_site *rsite;
if (lv_is_replicator_dev(lv))
dm_list_iterate_items(rsite, &first_seg(lv)->replicator->rsites) {
dm_list_iterate_items(rdev, &rsite->rdevices)
return rdev->replicator_dev->lv;
break;
}
return NULL;
}
/**
* Add VG open parameters to sorted cmd_vg list.
*
* Maintain the alphabeticaly ordered list, avoid duplications.
*
* \return Returns newly created or already present cmd_vg entry,
* or NULL in error case.
*/
struct cmd_vg *cmd_vg_add(struct dm_pool *mem, struct dm_list *cmd_vgs,
const char *vg_name, const char *vgid,
uint32_t flags)
{
struct cmd_vg *cvl, *ins;
if (!vg_name && !vgid) {
log_error("Either vg_name or vgid must be set.");
return NULL;
}
/* Is it already in the list ? */
if ((cvl = cmd_vg_lookup(cmd_vgs, vg_name, vgid)))
return cvl;
if (!(cvl = dm_pool_zalloc(mem, sizeof(*cvl)))) {
log_error("Allocation of cmd_vg failed.");
return NULL;
}
if (vg_name && !(cvl->vg_name = dm_pool_strdup(mem, vg_name))) {
dm_pool_free(mem, cvl);
log_error("Allocation of vg_name failed.");
return NULL;
}
if (vgid && !(cvl->vgid = dm_pool_strdup(mem, vgid))) {
dm_pool_free(mem, cvl);
log_error("Allocation of vgid failed.");
return NULL;
}
cvl->flags = flags;
if (vg_name)
dm_list_iterate_items(ins, cmd_vgs)
if (strcmp(vg_name, ins->vg_name) < 0) {
cmd_vgs = &ins->list; /* new position */
break;
}
dm_list_add(cmd_vgs, &cvl->list);
return cvl;
}
/**
* Find cmd_vg with given vg_name in cmd_vgs list.
*
* \param cmd_vgs List of cmd_vg entries.
*
* \param vg_name Name of VG to be found.
* \param vgid UUID of VG to be found.
*
* \return Returns cmd_vg entry if vg_name or vgid is found,
* NULL otherwise.
*/
struct cmd_vg *cmd_vg_lookup(struct dm_list *cmd_vgs,
const char *vg_name, const char *vgid)
{
struct cmd_vg *cvl;
dm_list_iterate_items(cvl, cmd_vgs)
if ((vgid && cvl->vgid && !strcmp(vgid, cvl->vgid)) ||
(vg_name && cvl->vg_name && !strcmp(vg_name, cvl->vg_name)))
return cvl;
return NULL;
}
/**
* Read and lock multiple VGs stored in cmd_vgs list alphabeticaly.
* On the success list head pointer is set to VGs' cmd_vgs.
* (supports FAILED_INCONSISTENT)
*
* \param cmd_vg Contains list of cmd_vg entries.
*
* \return Returns 1 if all VG in cmd_vgs list are correctly
* openned and locked, 0 otherwise.
*/
int cmd_vg_read(struct cmd_context *cmd, struct dm_list *cmd_vgs)
{
struct cmd_vg *cvl;
/* Iterate through alphabeticaly ordered cmd_vg list */
dm_list_iterate_items(cvl, cmd_vgs) {
2015-03-05 23:00:44 +03:00
cvl->vg = vg_read(cmd, cvl->vg_name, cvl->vgid, cvl->flags, 0);
if (vg_read_error(cvl->vg)) {
log_debug_metadata("Failed to vg_read %s", cvl->vg_name);
return 0;
}
cvl->vg->cmd_vgs = cmd_vgs; /* Make it usable in VG */
}
return 1;
}
/**
* Release opened and locked VGs from list.
*
* \param cmd_vgs Contains list of cmd_vg entries.
*/
2010-12-08 23:50:48 +03:00
void free_cmd_vgs(struct dm_list *cmd_vgs)
{
struct cmd_vg *cvl;
/* Backward iterate cmd_vg list */
dm_list_iterate_back_items(cvl, cmd_vgs) {
if (vg_read_error(cvl->vg))
release_vg(cvl->vg);
else
unlock_and_release_vg(cvl->vg->cmd, cvl->vg, cvl->vg_name);
cvl->vg = NULL;
}
}
/**
* Find all needed remote VGs for processing given LV.
* Missing VGs are added to VG's cmd_vg list and flag cmd_missing_vgs is set.
*/
int find_replicator_vgs(const struct logical_volume *lv)
{
struct replicator_site *rsite;
int ret = 1;
if (!lv_is_replicator_dev(lv))
return 1;
dm_list_iterate_items(rsite, &first_seg(lv)->replicator->rsites) {
if (!rsite->vg_name || !lv->vg->cmd_vgs ||
cmd_vg_lookup(lv->vg->cmd_vgs, rsite->vg_name, NULL))
continue;
ret = 0;
/* Using cmd memory pool for cmd_vg list allocation */
if (!cmd_vg_add(lv->vg->cmd->mem, lv->vg->cmd_vgs,
rsite->vg_name, NULL, 0)) {
lv->vg->cmd_missing_vgs = 0; /* do not retry */
stack;
break;
}
log_debug_metadata("VG: %s added as missing.", rsite->vg_name);
lv->vg->cmd_missing_vgs++;
}
return ret;
}
/**
* Read all remote VGs from lv's replicator sites.
* Function is used in activation context and needs all VGs already locked.
*/
int lv_read_replicator_vgs(const struct logical_volume *lv)
{
struct replicator_device *rdev;
struct replicator_site *rsite;
struct volume_group *vg;
if (!lv_is_replicator_dev(lv))
return 1;
dm_list_iterate_items(rsite, &first_seg(lv)->replicator->rsites) {
if (!rsite->vg_name)
continue;
2015-03-05 23:00:44 +03:00
vg = vg_read(lv->vg->cmd, rsite->vg_name, 0, 0, 0); // READ_WITHOUT_LOCK
if (vg_read_error(vg)) {
log_error("Unable to read volume group %s",
rsite->vg_name);
goto bad;
}
rsite->vg = vg;
/* FIXME: handling missing LVs needs to be better */
dm_list_iterate_items(rdev, &rsite->rdevices)
if (!(rdev->lv = find_lv(vg, rdev->name))) {
log_error("Unable to find %s in volume group %s",
rdev->name, rsite->vg_name);
goto bad;
}
}
return 1;
bad:
lv_release_replicator_vgs(lv);
return 0;
}
/**
* Release all VG resources taken by lv's replicator sites.
* Function is used in activation context and needs all VGs already locked.
*/
void lv_release_replicator_vgs(const struct logical_volume *lv)
{
struct replicator_site *rsite;
if (!lv_is_replicator_dev(lv))
return;
dm_list_iterate_back_items(rsite, &first_seg(lv)->replicator->rsites)
if (rsite->vg_name && rsite->vg) {
release_vg(rsite->vg);
rsite->vg = NULL;
}
}