Index: rmt.scm ================================================================== --- rmt.scm +++ rmt.scm @@ -78,16 +78,10 @@ #;(common:telemetry-log (conc "rmt:"(->string cmd)) payload: `((rid . ,rid) (params . ,params))) - ;;DOT digraph megatest_state_status { - ;;DOT ranksep=0; - ;;DOT // rankdir=LR; - ;;DOT node [shape="box"]; - ;;DOT "rmt:send-receive" -> MUTEXLOCK; - ;;DOT { edge [style=invis];"case 1" -> "case 2" -> "case 3" -> "case 4" -> "case 5" -> "case 6" -> "case 7" -> "case 8" -> "case 9" -> "case 10" -> "case 11"; } ;; do all the prep locked under the rmt-mutex (mutex-lock! *rmt-mutex*) ;; 1. check if server is started IFF cmd is a write OR if we are not on the homehost, store in runremote ;; 2. check the age of the connections. refresh the connection if it is older than timeout-20 seconds. @@ -97,60 +91,43 @@ (areapath *toppath*);; TODO - resolve from dbstruct to be compatible with multiple areas (runremote (or area-dat *runremote*)) (readonly-mode (rmtmod:calc-ro-mode runremote *toppath*))) - ;; DOT INIT_RUNREMOTE; // leaving off - doesn't really add to the clarity - ;; DOT MUTEXLOCK -> INIT_RUNREMOTE [label="no remote?"]; - ;; DOT INIT_RUNREMOTE -> MUTEXLOCK; ;; ensure we have a record for our connection for given area (if (not runremote) ;; can remove this one. should never get here. (begin (set! *runremote* (make-remote)) (set! runremote *runremote*))) ;; new runremote will come from this on next iteration - ;; DOT SET_HOMEHOST; // leaving off - doesn't really add to the clarity - ;; DOT MUTEXLOCK -> SET_HOMEHOST [label="no homehost?"]; - ;; DOT SET_HOMEHOST -> MUTEXLOCK; ;; ensure we have a homehost record (if (not (pair? (remote-hh-dat runremote))) ;; not on homehost (thread-sleep! 0.1) ;; since we shouldn't get here, delay a little (remote-hh-dat-set! runremote (common:get-homehost))) ;;(print "BB> readonly-mode is "readonly-mode" dbfile is "dbfile) (cond - ;;DOT EXIT; - ;;DOT MUTEXLOCK -> EXIT [label="> 15 attempts"]; {rank=same "case 1" "EXIT" } ;; give up if more than 15 attempts ((> attemptnum 15) (debug:print 0 *default-log-port* "ERROR: 15 tries to start/connect to server. Giving up.") (exit 1)) - ;;DOT CASE2 [label="local\nreadonly\nquery"]; - ;;DOT MUTEXLOCK -> CASE2; {rank=same "case 2" CASE2} - ;;DOT CASE2 -> "rmt:open-qry-close-locally"; ;; readonly mode, read request- handle it - case 2 ((and readonly-mode (member cmd api:read-only-queries)) (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 2") (rmt:open-qry-close-locally cmd 0 params) ) - ;;DOT CASE3 [label="write in\nread-only mode"]; - ;;DOT MUTEXLOCK -> CASE3 [label="readonly\nmode?"]; {rank=same "case 3" CASE3} - ;;DOT CASE3 -> "#f"; ;; readonly mode, write request. Do nothing, return #f (readonly-mode (extras-readonly-mode *rmt-mutex* *default-log-port* cmd params)) ;; This block was for pre-emptively resetting the connection if there had been no communication for some time. ;; I don't think it adds any value. If the server is not there, just fail and start a new connection. ;; also, the expire-time calculation might not be correct. We want, time-since-last-server-access > (server:get-timeout) ;; - ;;DOT CASE4 [label="reset\nconnection"]; - ;;DOT MUTEXLOCK -> CASE4 [label="have connection,\nlast_access > expire_time"]; {rank=same "case 4" CASE4} - ;;DOT CASE4 -> "rmt:send-receive"; ;; reset the connection if it has been unused too long ((and runremote (remote-conndat runremote) (> (current-seconds) ;; if it has been more than server-timeout seconds since last contact, close this connection and start a new on (+ (http-transport:server-dat-get-last-access (remote-conndat runremote)) @@ -159,25 +136,19 @@ (http-transport:close-connections area-dat: runremote) (remote-conndat-set! runremote #f) ;; invalidate the connection, thus forcing a new connection. (mutex-unlock! *rmt-mutex*) (rmt:send-receive-orig cmd rid params attemptnum: attemptnum)) - ;;DOT CASE5 [label="local\nread"]; - ;;DOT MUTEXLOCK -> CASE5 [label="server not required,\non homehost,\nread-only query"]; {rank=same "case 5" CASE5}; - ;;DOT CASE5 -> "rmt:open-qry-close-locally"; ;; on homehost and this is a read ((and (not (remote-force-server runremote)) ;; honor forced use of server, i.e. server NOT required (cdr (remote-hh-dat runremote)) ;; on homehost (member cmd api:read-only-queries)) ;; this is a read (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 5") (rmt:open-qry-close-locally cmd 0 params)) - ;;DOT CASE6 [label="init\nremote"]; - ;;DOT MUTEXLOCK -> CASE6 [label="on homehost,\nwrite query,\nhave server,\ncan't reach it"]; {rank=same "case 6" CASE6}; - ;;DOT CASE6 -> "rmt:send-receive"; ;; on homehost and this is a write, we already have a server, but server has died ((and (cdr (remote-hh-dat runremote)) ;; on homehost (not (member cmd api:read-only-queries)) ;; this is a write (remote-server-url runremote) ;; have a server (not (server:ping (remote-server-url runremote)))) ;; server has died. NOTE: this is not a cheap call! Need better approach. @@ -185,25 +156,19 @@ (remote-force-server-set! runremote (common:force-server?)) (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 6") (rmt:send-receive-orig cmd rid params attemptnum: attemptnum)) - ;;DOT CASE7 [label="homehost\nwrite"]; - ;;DOT MUTEXLOCK -> CASE7 [label="server not required,\non homehost,\na write,\nhave a server"]; {rank=same "case 7" CASE7}; - ;;DOT CASE7 -> "rmt:open-qry-close-locally"; ;; on homehost and this is a write, we already have a server ((and (not (remote-force-server runremote)) ;; honor forced use of server, i.e. server NOT required (cdr (remote-hh-dat runremote)) ;; on homehost (not (member cmd api:read-only-queries)) ;; this is a write (remote-server-url runremote)) ;; have a server (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 4.1") (rmt:open-qry-close-locally cmd 0 params)) - ;;DOT CASE8 [label="force\nserver"]; - ;;DOT MUTEXLOCK -> CASE8 [label="server not required,\nhave homehost info,\nno connection yet,\nnot a read-only query"]; {rank=same "case 8" CASE8}; - ;;DOT CASE8 -> "rmt:open-qry-close-locally"; ;; on homehost, no server contact made and this is a write, passively start a server ((and (not (remote-force-server runremote)) ;; honor forced use of server, i.e. server NOT required (cdr (remote-hh-dat runremote)) ;; have homehost (not (remote-server-url runremote)) ;; no connection yet (not (member cmd api:read-only-queries))) ;; not a read-only query @@ -217,13 +182,10 @@ (remote-force-server-set! runremote (common:force-server?)) (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 8.1") (rmt:open-qry-close-locally cmd 0 params)) - ;;DOT CASE9 [label="force server\nnot on homehost"]; - ;;DOT MUTEXLOCK -> CASE9 [label="no connection\nand either require server\nor not on homehost"]; {rank=same "case 9" CASE9}; - ;;DOT CASE9 -> "start\nserver" -> "rmt:send-receive"; ((or (and (remote-force-server runremote) ;; we are forcing a server and don't yet have a connection to one (not (remote-conndat runremote))) (and (not (cdr (remote-hh-dat runremote))) ;; not on a homehost (not (remote-conndat runremote)))) ;; and no connection (debug:print-info 12 *default-log-port* "rmt:send-receive, case 9, hh-dat: " (remote-hh-dat runremote) " conndat: " (remote-conndat runremote)) @@ -231,27 +193,19 @@ (if (not (server:check-if-running *toppath*)) ;; who knows, maybe one has started up? (server:start-and-wait *toppath*)) (remote-conndat-set! runremote (rmt:get-connection-info *toppath*)) ;; calls client:setup which calls client:setup-http (rmt:send-receive-orig cmd rid params attemptnum: attemptnum)) ;; TODO: add back-off timeout as - ;;DOT CASE10 [label="on homehost"]; - ;;DOT MUTEXLOCK -> CASE10 [label="server not required,\non homehost"]; {rank=same "case 10" CASE10}; - ;;DOT CASE10 -> "rmt:open-qry-close-locally"; ;; all set up if get this far, dispatch the query ((and (not (remote-force-server runremote)) (cdr (remote-hh-dat runremote))) ;; we are on homehost (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 10") (rmt:open-qry-close-locally cmd (if rid rid 0) params)) - ;;DOT CASE11 [label="send_receive"]; - ;;DOT MUTEXLOCK -> CASE11 [label="else"]; {rank=same "case 11" CASE11}; - ;;DOT CASE11 -> "rmt:send-receive" [label="call failed"]; - ;;DOT CASE11 -> "RESULT" [label="call succeeded"]; ;; not on homehost, do server query (else (extras-case-11 *default-log-port* runremote cmd params attemptnum rid))))) - ;;DOT } ;; bunch of small functions factored out of send-receive to make debug easier ;; (define (extras-case-11 *default-log-port* runremote cmd params attemptnum rid)