Commit Graph

78 Commits

Author SHA1 Message Date
Aliaksandr Valialkin
02b2064d8e app/vminsert/netstorage: do not increment vm_rpc_rows_lost_total when all the vmstorage nodes are unavailable, since vminsert retries sending the data instead of dropping it 2020-05-28 22:36:56 +03:00
Aliaksandr Valialkin
7a61357b5d app/vminsert/netstorage: make sure that the the data is always replicated among -replicationFactor vmstorage nodes
Previously vminsert could write multiple copies of the data to a single vmstorage node when the ingestion rate
exceeds the maximum throughput for connections to vmstorage nodes.
2020-05-28 19:59:07 +03:00
Aliaksandr Valialkin
77e5165e7b app/vminsert: add -replicationFactor command-line flag for enabling data replication among available -storageNode instances 2020-05-27 17:29:44 +03:00
Aliaksandr Valialkin
b4e3bffe4b app/vminsert/netstorage: emit warnings instead of errors when re-routing data to healthy storage nodes 2020-05-27 16:31:41 +03:00
Aliaksandr Valialkin
75f2f3b09d app/vminsert/netstorage: improve ingestion performance when a single vmstorage node is slower than other vmstorage nodes
Previously the ingestion performance has been limited by the slowest vmstorage node.
Now vminsert should re-route data from the slowest vmstorage node to the remaining nodes.
2020-05-27 15:08:22 +03:00
Aliaksandr Valialkin
9844845d79 app/vminsert: tune the maximum summary buffer size for pending data to 1/4 of available RAM, since 1/2 of RAM is too big considering GOGC overhead 2020-05-25 02:00:37 +03:00
Aliaksandr Valialkin
4a82631e44 app/vminsert: limit the summary buffer sizes for all the storage nodes to a half of the allowed memory 2020-05-25 01:39:33 +03:00
Aliaksandr Valialkin
4bd3d4b148 app/vminsert/netstorage: do not return error from storageNode.flushBufLocked when the buffer has been successfully re-routed to healthy nodes
This should reduce the number of false errors in the log and the number of falsely lost rows
2020-05-22 18:29:43 +03:00
Aliaksandr Valialkin
6edc33d9bb app/vminsert/netstorage: capture the first error instead of the last error when sending data to vmstorage
The first error has more chances to point to the real root cause of the issue.
2020-05-22 17:49:33 +03:00
Aliaksandr Valialkin
716bbe79d4 app/vminsert/netstorage: increase timeout for waiting for ack message after sending big data block to vmstorage 2020-04-28 11:19:46 +03:00
Aliaksandr Valialkin
989d84cf3f app/{vminsert,vmstorage}: wait for ack from vmstorage after each packet sent to it from vminsert
This should protect from possible data loss when `vmstorage` is stopped while the packet is sent from `vminsert`.

This commit switches to new protocol between vminsert and vmstorage, which is incompatible
with the previous protocol. So it is required that both vminsert and vmstorage nodes are updated.
2020-04-27 09:53:26 +03:00
Aliaksandr Valialkin
7ee7614e90 app/vmagent: initial implementation for vmagent 2020-02-23 17:31:54 +02:00
Aliaksandr Valialkin
347aaba79d lib/{storage,mergeset}: use time.Ticker instead of time.Timer where appropriate
It has been appeared that time.Timer was used in places where time.Ticker must be used instead.
This could result in blocked goroutines as in the https://github.com/VictoriaMetrics/VictoriaMetrics/issues/316 .
2020-02-13 13:21:48 +02:00
Aliaksandr Valialkin
4d70a81e18 app/vminsert: do not drop pending rows if all the vmstorage backends are unavailable
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/294
2020-01-24 22:10:10 +02:00
Aliaksandr Valialkin
ecddba30fe app/vminsert/netstorage: increase timeout for pushing data from vminsert to vmstorage by 3x
Our clients report that the previous timeout could lead to frequent errors when
vmstorage starts background merge for big parts on slow HDD.
2020-01-21 18:21:49 +02:00
Aliaksandr Valialkin
c81a89a8ed app/vminsert: add /api/v1/import handler
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6
2019-12-09 22:37:49 +02:00
Aliaksandr Valialkin
8d398af92f app/vminsert/netstorage: mention the data size that cannot be sent to vmstorage
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/175
2019-09-25 12:53:41 +03:00
Aliaksandr Valialkin
73ac7b8dd6 app/vminsert/netstorage: make sure the conn exists before closing it in storageNode.closeBrokenConn
The conn can be missing or already closed during the call to storageNode.closeBrokenConn.
Prevent `nil pointer dereference` panic by verifying whether the conn is already closed.

Thanks to @CH-anhngo for reporting the issue.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/189
2019-09-25 10:36:50 +03:00
Aliaksandr Valialkin
550a12415a app/vminsert/netstorage: log network errors when sending data to vmstorage nodes
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/175
2019-09-13 22:26:24 +03:00
Aliaksandr Valialkin
ed50b8792b app/vminsert/netstorage: reduce the maximum buffer size for rerouted rows, so it occupies less RAM 2019-09-11 14:50:30 +03:00
Aliaksandr Valialkin
2f4c950fe9 app/vminsert/netstorage: send per-storageNode bufs to vmstorage nodes in parallel
This should improve the maximum ingestion throughput.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/175
2019-09-11 14:50:19 +03:00
Aliaksandr Valialkin
694cc59ed1 app/vminsert/netstorage: dynamically adjust timeouts for sending packets from vminsert to vmstorage depending on packet size
Bigger packets will have more chances to be sent to vmstorage.
2019-09-11 14:50:14 +03:00
Aliaksandr Valialkin
697de90893 app/vminsert: do not drop data in reroutedBuf if all the storage nodes are unhealthy 2019-08-23 10:38:19 +03:00
Aliaksandr Valialkin
a5dc54efc3 app/vminsert: properly limit the size of reroutedBuf 2019-08-23 10:29:51 +03:00
Aliaksandr Valialkin
c197641978 all: return 503 http error if service is temporarily unavailable
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/156
2019-08-23 09:49:50 +03:00
Aliaksandr Valialkin
5fcdb4a59a app/vminsert: improve handling of unhealthy vmstorage nodes
* Spread load evenly among remaining healthy nodes instead of hammering
  the next node after the unhealthy node.
* Make sure that the packet is flushed to storage node before returning success.
  Previously packets could stay in local buffers and thus lost on connection errors.
* Keep rows in the limited memory when all the storage nodes are unhealthy.
2019-06-09 00:42:36 +03:00
Aliaksandr Valialkin
f37903adb3 app/vminsert: add -rpc.disableCompression command-line flag for reducing CPU usage at the cost of higher network bandwidth usage 2019-05-24 12:51:07 +03:00
Aliaksandr Valialkin
24578b4bb1 all: open-sourcing cluster version 2019-05-23 00:25:38 +03:00