Support #660

When system-v service fails to advertise doe to max servers per service is full shall return error

Added by Madars almost 3 years ago. Updated almost 3 years ago.

Status:NewStart date:03/28/2021
Priority:Normal (Code 4)Due date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

This situation later causes in System-V mode to remove process RQ addr:

Code block:

        /* Register stuff in shared memory! */
        if (use_sem)
        {
#ifdef EX_USE_SYSVQ
            ret=ndrx_shm_install_svc(entry->svc_nm, 0, ndrx_epoll_resid_get());
#else
            ret=ndrx_shm_install_svc(entry->svc_nm, 0, G_server_conf.srv_id);
#endif
        }

        /* Release semaphore! */
        if (use_sem) ndrx_unlock_svc_op(__func__);
        /* ###################### CRITICAL SECTION, END ########################## */

        if (EXSUCCEED!=ret)
        {
            NDRX_LOG(log_error, "Service shared memory full - currently ignore error!");
            ret=EXSUCCEED;
        }

Only needs to analyse how it would correlate with epoll mode, is it used at all there? Or only affected for poll / system-v modes?

History

#1 Updated by Madars almost 3 years ago

Another issue with System V - even NDRX_SVCMAX is set to large number and booting the same large number of servers (for example 3000), under psvc -r only 512 resources appear. Others are left empty.


$ xadmin psvc -r

  SLOT SERVICE NAME NSRV FLAGS CSRVS TCLST CMAX CNODES
------ ------------ ---- ----- ----- ----- ---- --------------------------------
 18210 EXBENCH        4K     1     0     0    0 00000000000000000000000000000000

        RES NO IDENTIFIER SERVERS
        ------ ---------- -------

           506    2099140       1
           507    2099131       1
           508    2099122       1
           509    2099113       1
           510    2099104       1
           511    2099095       1
           512 1667789686       0
           513          0       0
           514          0       0
           515          0       0
           516          0       0
           517          0       0
           518          0       0
           519          0       0
           520          0       0
           521          0       0
           522          0       0
           523          0       0

#2 Updated by Lauris almost 3 years ago

Issue with:

           512 1667789686       0
           513          0       0

is related with fact of the:

typedef struct
{
    command_reply_t rply;
    int slot;                       /**< Position in SHM                      */
    char service[MAXTIDENT+1];      /**< name of the service                  */
    int srvs;                       /**< Count of servers advertising this service*/
    int flags;                      /**< service flags                        */

    int csrvs;                      /**< Number of advertises in cluster      */
    int totclustered;               /**< Total clustered nodes                */
    int cnodes_max_id;              /**< Max id of cluster nodes in list (for fast search) */
    cnodeinfo_t cnodes[CONF_NDRX_NODEID_COUNT]; /**< List of cluster nodes */
    /**< Server ID (fixed number xadmin output) */
    ndrx_shm_resid_t resids[CONF_NDRX_MAX_SRVIDS_XADMIN];
    int resnr;                                  /**< Number of local resources              */

} command_reply_shm_psvc_t;

CONF_NDRX_MAX_SRVIDS_XADMIN constant. Probably needs to be reworked to have resids dynamically allocated for full listing.

Also available in: Atom PDF