4、tracker的三个任务函数之一tracker_mem_check_alive

一、概要

前文提到调度线程sched_thread_entrance会调度以下三个任务函数log_sync_func、tracker_mem_check_alive、tracker_write_status_to_file。其中log_sync_func用于将缓存中的log刷新到磁盘中，它的使用与log_init()紧密相连，且log_to_cache参数用于设置是否不经过缓存直接刷新到磁盘；tracker_write_status_to_file向文件data/.tracker_status中写入TRACKER_STATUS_ITEM_UP_TIME、TRACKER_STATUS_ITEM_LAST_CHECK_TIME对应的时间。本文不讨论这两个函数。

二、tracker_mem_check_alive

通过时间对比，找出不活动的 storage server，接下来并更改 storage 的状态为 FDFS_STORAGE_STATUS_OFFLINE。

int tracker_mem_check_alive(void *arg)
{
	FDFSStorageDetail **ppServer;
	FDFSStorageDetail **ppServerEnd;
	FDFSGroupInfo **ppGroup;
	FDFSGroupInfo **ppGroupEnd;
	FDFSStorageDetail *deactiveServers[FDFS_MAX_SERVERS_EACH_GROUP];
	int deactiveCount;
	time_t current_time;

	current_time = g_current_time;
	ppGroupEnd = g_groups.groups + g_groups.count;
	for (ppGroup=g_groups.groups; ppGroup<ppGroupEnd; ppGroup++) //对每个组
	{
	deactiveCount = 0;
	ppServerEnd = (*ppGroup)->active_servers + (*ppGroup)->active_count;
	for (ppServer=(*ppGroup)->active_servers; ppServer<ppServerEnd; ppServer++)  //对组中的每个storage server
	{
		if (current_time - (*ppServer)->stat.last_heart_beat_time > \
			g_check_active_interval)   //本例配置为120s
		{
			deactiveServers[deactiveCount] = *ppServer;
			deactiveCount++;
			if (deactiveCount >= FDFS_MAX_SERVERS_EACH_GROUP)
			{
				break;
			}
		}
	}

	if (deactiveCount == 0)
	{
		continue;
	}

	ppServerEnd = deactiveServers + deactiveCount;
	for (ppServer=deactiveServers; ppServer<ppServerEnd; ppServer++)
	{
		(*ppServer)->status = FDFS_STORAGE_STATUS_OFFLINE;   //对于超时无心跳的storage server。
		tracker_mem_deactive_store_server(*ppGroup, *ppServer); //更新对应组的信息
		if (g_use_storage_id)
		{ //log
		}
		else
		{
		}
	}
	}

	if ((!g_if_leader_self) || (!g_if_use_trunk_file))
	{
		return 0;
	}

	for (ppGroup=g_groups.groups; ppGroup<ppGroupEnd; ppGroup++)
	{
	if ((*ppGroup)->pTrunkServer != NULL)  
	{//此处处理暂时不考虑
	}
	else
	{
		tracker_mem_find_trunk_server(*ppGroup, true);
	}
	}

	return 0;
}

三、Group与storage server的数据信息

FDFSGroups g_groups（全局变量）------FDFSGroupInfo-------FDFSStorageDetail：

typedef struct
{
	int alloc_size;   //alloc group count
	int count;  //group count
	FDFSGroupInfo **groups;
	FDFSGroupInfo **sorted_groups; //groups order by group_name
	FDFSGroupInfo *pStoreGroup;  //the group to store uploaded files
	int current_write_group;  //current group index to upload file
	byte store_lookup;  //store to which group, from conf file
	byte store_server;  //store to which storage server, from conf file
	byte download_server; //download from which storage server, from conf file
	byte store_path;  //store to which path, from conf file
	char store_group[FDFS_GROUP_NAME_MAX_LEN + 8];  //for 8 bytes aliginment
} FDFSGroups;

typedef struct   //该结构体对应每一个组
{
	char group_name[FDFS_GROUP_NAME_MAX_LEN + 8];   //for 8 bytes alignment
	int64_t total_mb;  //total disk storage in MB
	int64_t free_mb;  //free disk storage in MB
	int64_t trunk_free_mb;  //trunk free disk storage in MB
	int alloc_size;  //alloc storage count
	int count;    //total server count
	int active_count; //active server count
	int storage_port;  //storage server port
	int storage_http_port; //storage http server port
	int current_trunk_file_id;  //current trunk file id report by storage
	FDFSStorageDetail **all_servers;   //all storage servers
	FDFSStorageDetail **sorted_servers;  //storages order by ip addr
	FDFSStorageDetail **active_servers;  //storages order by ip addr
	FDFSStorageDetail *pStoreServer;  //for upload priority mode
	FDFSStorageDetail *pTrunkServer;  //point to the trunk server
	char last_trunk_server_id[FDFS_STORAGE_ID_MAX_SIZE];

#ifdef WITH_HTTPD
	FDFSStorageDetail **http_servers;  //storages order by ip addr
	int http_server_count; //http server count
	int current_http_server; //current http server index
#endif

	int current_read_server;   //current read storage server index
	int current_write_server;  //current write storage server index

	int store_path_count;  //store base path count of each storage server

	/* subdir_count * subdir_count directories will be auto created
	   under each store_path (disk) of the storage servers
	*/
	int subdir_count_per_path;

	int **last_sync_timestamps;//row for src storage, col for dest storage

	int chg_count;   //current group changed count
	int trunk_chg_count;   //trunk server changed count
	time_t last_source_update;  //last source update timestamp
	time_t last_sync_update;    //last synced update timestamp
} FDFSGroupInfo;

typedef struct StructFDFSStorageDetail //该结构体反映了每一个storage server。
{
	char status;
	char padding;  //just for padding
	char id[FDFS_STORAGE_ID_MAX_SIZE];
	char ip_addr[IP_ADDRESS_SIZE];
	char version[FDFS_VERSION_SIZE];
	char domain_name[FDFS_DOMAIN_NAME_MAX_SIZE];

	struct StructFDFSStorageDetail *psync_src_server;
	int64_t *path_total_mbs; //total disk storage in MB
	int64_t *path_free_mbs;  //free disk storage in MB

	int64_t total_mb;  //total disk storage in MB
	int64_t free_mb;  //free disk storage in MB
	int64_t changelog_offset;  //changelog file offset

	time_t sync_until_timestamp;
	time_t join_time;  //storage join timestamp (create timestamp)
	time_t up_time;    //startup timestamp

	int store_path_count;  //store base path count of each storage server
	int subdir_count_per_path;
	int upload_priority; //storage upload priority

	int storage_port;   //storage server port
	int storage_http_port; //storage http server port

	int current_write_path; //current write path index

	int chg_count;    //current server changed counter
	int trunk_chg_count;   //trunk server changed count
	FDFSStorageStat stat;

#ifdef WITH_HTTPD
	int http_check_last_errno;
	int http_check_last_status;
	int http_check_fail_count;
	char http_check_error_info[256];
#endif
} FDFSStorageDetail

四、(*ppServer)->stat.last_heart_beat_time
storage有个专门的report线程向tracker server发送报告，包括发送TRACKER_PROTO_CMD_STORAGE_BEAT至tracker。tracker在tracker_deal_task()中会调用tracker_deal_storage_beat(pTask)（参考《3 tracker中的recv_notify_read》），更新心跳时间。