| " "hostId | " "host ip | " "shard | " "mirror | " // mirror # within the shard // i don't remember the last time i used this, so let's // just comment it out to save space //"group mask | " //"ip1 | " //"ip2 | " //"udp port | " // this is now more or less obsolete //"priority udp port | " //"dns client port | " "http port | " // this is now obsolete since ide channel is. it was used // so that only the guy with the token could merge, // and it made sure that only one merge per ide channel // and per group was going on at any one time for performance // reasons. //"token group | " //"best switch id | " //"actual switch id | " //"switch id | " // this is now fairly obsolete //"ide channel | " //"HD temps (C) | " "GB version | " //"resends sent | " //"errors recvd | " //"ETRYAGAINS recvd | " "" "dgrams resent | " "" "errors recvd | " "" "ETRY AGAINS recvd | " "" "dgrams to | " "" "dgrams from | " //"" //"loadavg | " "" "avg split time | " "splits done | " "" "status | " "" "slow reads | " "docs indexed | " "" "mem used | " "" "cpu used | " "" "disk used | " "" "max ping1 | " "" "ping1 age | " //"ip1 | " "" "ping1 | " "%s"// "ip2 | " //"inSync | ", //"avg roundtrip | " //"std. dev. | note | ", TABLE_STYLE , colspan , cs, sort, DARK_BLUE , cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, cs, shotcol ); // loop through each host we know and print it's stats long nh = g_hostdb.getNumHosts(); // should we reset resends, errorsRecvd and ETRYAGAINS recvd? if ( r->getLong("reset",0) ) { for ( long i = 0 ; i < nh ; i++ ) { // get the ith host (hostId) Host *h = g_hostdb.getHost ( i ); h->m_totalResends = 0; h->m_errorReplies = 0; h->m_etryagains = 0; h->m_dgramsTo = 0; h->m_dgramsFrom = 0; } } // sort hosts if needed long hostSort [ MAX_HOSTS ]; for ( long i = 0 ; i < nh ; i++ ) hostSort [ i ] = i; switch ( sort ) { case 1: gbsort ( hostSort, nh, sizeof(long), pingSort1 ); break; case 2: gbsort ( hostSort, nh, sizeof(long), pingSort2 ); break; case 3: gbsort ( hostSort, nh, sizeof(long), resendsSort ); break; case 4: gbsort ( hostSort, nh, sizeof(long), errorsSort ); break; case 5: gbsort ( hostSort, nh, sizeof(long), tryagainSort ); break; case 6: gbsort ( hostSort, nh, sizeof(long), dgramsToSort ); break; case 7: gbsort ( hostSort, nh, sizeof(long), dgramsFromSort ); break; //case 8: gbsort ( hostSort, nh, sizeof(long), loadAvgSort ); break; case 9: gbsort ( hostSort, nh, sizeof(long), memUsedSort ); break; case 10:gbsort ( hostSort, nh, sizeof(long), cpuUsageSort ); break; case 11:gbsort ( hostSort, nh, sizeof(long), pingAgeSort ); break; case 12:gbsort ( hostSort, nh, sizeof(long), flagSort ); break; case 13:gbsort ( hostSort, nh, sizeof(long), splitTimeSort ); break; case 14:gbsort ( hostSort, nh, sizeof(long), pingMaxSort ); break; case 15:gbsort ( hostSort, nh, sizeof(long), slowDiskSort ); break; case 16:gbsort ( hostSort, nh, sizeof(long), defaultSort ); break; case 17:gbsort ( hostSort, nh, sizeof(long), diskUsageSort ); break; } // we are the only one that uses these flags, so set them now /* static char s_properSet = 0; if ( ! s_properSet ) { s_properSet = 1; g_hostdb.setOnProperSwitchFlags(); } */ if ( format == FORMAT_XML ) { sb.safePrintf("
| %li | " "%s | " // hostname "%li | " // group "%li | " // stripe //"0x%08lx | " // group mask //"%s | " // ip1 //"%s | " // ip2 //"%hi | " // port //"%hi | " // client port "%hi | " // http port //"%li | " // token group num //"%li | " // switch group //"%s | " // tmpN //"%li | " // ide channel // hd temps // no, this is gb version now "%li | " // error replies "%li | " // etryagains "%li | " // # dgrams sent to "%lli | " // # dgrams recvd from "%lli | " // loadavg //"%.2f | " // split time "%li | " // splits done "%li | " // flags "%s | " // slow disk reads "%li | " // docs indexed "%li | " // percent mem used "%s%.1f%%%s | " // cpu usage "%.1f%% | " // disk usage "%s | " // ping max "%s | " // ping age "%lims | " // ping "%s | " //"%s | " //"%lims | " "%s | " "
| spareId | " "host name | " "ip1 | " "ip2 | " //"udp port | " //"priority udp port | " //"dns client port | " "http port | " //"switch id | " // this is now fairly obsolete //"ide channel | " "note | ", TABLE_STYLE, DARK_BLUE ); for ( long i = 0; i < g_hostdb.m_numSpareHosts; i++ ) { // get the ith host (hostId) Host *h = g_hostdb.getSpare ( i ); char ipbuf1[64]; char ipbuf2[64]; strcpy(ipbuf1,iptoa(h->m_ip)); strcpy(ipbuf2,iptoa(h->m_ipShotgun)); // print it sb.safePrintf ( "
| %li | " "%s | " "%s | " "%s | " //"%hi | " //"%hi | " // priority udp port //"%hi | " "%hi | " //"%i | " // switch id //"%li | " // ide channel "%s | " "
| proxyId | " "type | " "host name | " "ip1 | " "ip2 | " //"udp port | " //"priority udp port | " //"dns client port | " "http port | " //"switch id | " "max ping1 | " "ping1 age | " "ping1 | " //"ping2 | " // this is now fairly obsolete //"ide channel | " "note | ", TABLE_STYLE, DARK_BLUE ); for ( long i = 0; i < g_hostdb.m_numProxyHosts; i++ ) { // get the ith host (hostId) Host *h = g_hostdb.getProxy ( i ); char ptr[256]; long pingAge = generatePingMsg(h, nowmsLocal, ptr); char ipbuf1[64]; char ipbuf2[64]; strcpy(ipbuf1,iptoa(h->m_ip)); strcpy(ipbuf2,iptoa(h->m_ipShotgun)); // host can have 2 ip addresses, get the one most // similar to that of the requester long eip = g_hostdb.getBestIp ( h , s->m_ip ); char ipbuf3[64]; strcpy(ipbuf3,iptoa(eip)); char pms[64]; if ( h->m_pingMax < 0 ) sprintf(pms,"???"); else sprintf(pms,"%lims",h->m_pingMax); // the sync status ascii-ized char *type = "proxy"; if ( h->m_type == HT_QCPROXY ) type = "qcproxy"; if ( h->m_type == HT_SCPROXY ) type = "scproxy"; // print it sb.safePrintf ( "
| " "%li | " "%s | " "%s | " "%s | " "%s | " //"%hi | " //"%hi | " // priority udp port //"%hi | " "%hi | " //"%i | " // switch id "%s | " // ping max "%ldms | " // ping age "%s | " // ping //"%li | " // ide channel "%s | " "|
| host ip | " "The primary IP address of the host." " | " "||||||||
| shard | " "" "The index is split into shards. Which shard does this " "host serve?" " | " "||||||||
| mirror | " "" "A shard can be mirrored multiple times for " "data redundancy." " | " "||||||||
| ip2 | " "The secondary IP address of the host." " | " "||||||||
| udp port | " "The UDP port the host uses to send and recieve " "datagrams." " | " "||||||||
| dns client port | " "The UDP port used to send and receive dns traffic with." " | " "||||||||
| http port | " "The port you can connect a browser to." " | " "||||||||
| best switch id | " "The host prefers to be on this switch because it " "needs to send a lot of data to other hosts on this swtich. " "Therefore, ideally, the best switch id should match the " "actual switch id for optimal performance." " | " "||||||||
| switch id | " "Hosts that share the same switch id are " "physically on the same switch." " | " "||||||||
| dgrams resent | " "How many datagrams have had to be resent to a host " "because it was not ACKed quick enough or because it was " "fully ACKed but the entire request was resent in case " "the host was reset." " | " "||||||||
| errors recvd | " "How many errors were received from a host in response " "to a request to retrieve or insert data." " | " "||||||||
| ETRYAGAINS recvd | " "How many ETRYAGAIN were received in response to a " "request to add data. Usually because the host's memory " "is full and it is dumping its data to disk. This number " "can be high if the host if failing to dump the data " "to disk because of some malfunction, and it can therefore " "bottleneck the entire cluster." " | " "||||||||
| dgrams to | " "How many datagrams were sent to the host from the " "selected host since startup. Includes ACK datagrams. This " "can actually be higher than the number of dgrams read " "when the selected host is the same as the host in the " "table because of resends. Gigablast will resend datagrams " "that are not promptly ACKknowledged." " | " "||||||||
| dgrams from | " "How many datagrams were received from the host by the " "selected host since startup. Includes ACK datagrams." " | " "||||||||
| avg split time | " "Average time this host took to compute the docids " "for a query. Useful for guaging the slowness of a host " "compare to other hosts." " | " "||||||||
| splits done | " "Number of queries this host completed. Used in " "computation of the avg split time." " | " "||||||||
| status | " "Status flags for the host. See key below." " | " "||||||||
| slow reads | " "Number of slow disk reads the host has had. " "When this is big compared to other hosts it is a good " "indicator its drives are relatively slow." " | " "||||||||
| docs indexed | " "Number of documents this host has indexed over all " "collections. All hosts should have close to the same " "number in a well-sharded situation." " | " "||||||||
| loadavg | " //"1-minute sliding-window load average from " //"/proc/loadavg." //" | " //"||||||||
| mem used | " "Percentage of memory currently used." " | " "||||||||
| cpu used | " "Percentage of cpu resources in use by the gb process." " | " "||||||||
| disk used | " "Percentage of disk in use. When this gets close to " "100%% you need to do something." " | " "||||||||
| max ping1 | " "The worst ping latency from host to host." " | " "||||||||
| ping1 age | " "How long ago the last ping request was sent to " "this host. Let's us know how fresh the ping time is." " | " "||||||||
| ping1 | " "Ping time to this host on the primary network." " | " "||||||||
| ping2 | " "Ping time to this host on the seconday/shotgun " "network. This column is not visible if the shotgun " "network is not enabled in the master controls." " | " "||||||||
| M (status flag) | " "Indicates host is merging files on disk." " | " "||||||||
| D (status flag) | " "Indicates host is dumping data to disk." " | " "||||||||
| S (status flag) | " "Indicates host has outstanding spiders." " | " "||||||||
| y (status flag) | " "Indicates host is performing the daily merge." " | " "||||||||
| R (status flag) | " "Indicates host is performing a rebalance operation." " | " "||||||||
| F (status flag) | " "Indicates host has foreign records and requires " "a rebalance operation." " | " "||||||||
| x (status flag) | " "Indicates host has abruptly exited due to a fatal " "error (cored) and " "restarted itself." " | " "||||||||