mirror of
https://github.com/overleaf/overleaf.git
synced 2024-11-21 20:47:08 -05:00
Merge pull request #64 from overleaf/sk-skip-failing-projects-in-swap-job
Skip failing projects in swap job, and give up after a certain number
This commit is contained in:
commit
b6fc48645a
1 changed files with 30 additions and 2 deletions
|
@ -14,6 +14,7 @@ import java.io.InputStream;
|
|||
import java.sql.Timestamp;
|
||||
import java.time.Duration;
|
||||
import java.time.LocalDateTime;
|
||||
import java.util.ArrayList;
|
||||
import java.util.Timer;
|
||||
import java.util.concurrent.atomic.AtomicInteger;
|
||||
|
||||
|
@ -105,6 +106,8 @@ public class SwapJobImpl implements SwapJob {
|
|||
}
|
||||
|
||||
private void doSwap_() {
|
||||
ArrayList<String> exceptionProjectNames = new ArrayList<String>();
|
||||
|
||||
Log.info("Running swap number {}", swaps.get() + 1);
|
||||
long totalSize = repoStore.totalSize();
|
||||
Log.info("Size is {}/{} (high)", totalSize, highWatermarkBytes);
|
||||
|
@ -114,15 +117,40 @@ public class SwapJobImpl implements SwapJob {
|
|||
return;
|
||||
}
|
||||
int numProjects = dbStore.getNumProjects();
|
||||
// while we have too many projects on disk
|
||||
while (
|
||||
(totalSize = repoStore.totalSize()) > lowWatermarkBytes &&
|
||||
(numProjects = dbStore.getNumUnswappedProjects()) > minProjects
|
||||
) {
|
||||
// check if we've had too many exceptions so far
|
||||
if (exceptionProjectNames.size() >= 20) {
|
||||
StringBuilder sb = new StringBuilder();
|
||||
for (String s: exceptionProjectNames) {
|
||||
sb.append(s);
|
||||
sb.append(' ');
|
||||
}
|
||||
Log.error(
|
||||
"Too many exceptions while running swap, giving up on this run: {}",
|
||||
sb.toString()
|
||||
);
|
||||
break;
|
||||
}
|
||||
// get the oldest project and try to swap it
|
||||
String projectName = dbStore.getOldestUnswappedProject();
|
||||
try {
|
||||
evict(projectName);
|
||||
} catch (IOException e) {
|
||||
Log.warn("[{}] Exception while swapping, giving up", projectName, e);
|
||||
} catch (Throwable t) {
|
||||
Log.warn("[{}] Exception while swapping, mark project and move on", projectName, t);
|
||||
// NOTE: this is something of a hack. If a project fails to swap we get stuck in a
|
||||
// loop where `dbStore.getOldestUnswappedProject()` gives the same failing project over and over again,
|
||||
// which fills up the disk with errors. By touching the access time we can mark the project as a
|
||||
// non-candidate for swapping. Ideally we should be checking the logs for these log events and fixing
|
||||
// whatever is wrong with the project
|
||||
dbStore.setLastAccessedTime(
|
||||
projectName,
|
||||
Timestamp.valueOf(LocalDateTime.now())
|
||||
);
|
||||
exceptionProjectNames.add(projectName);
|
||||
}
|
||||
}
|
||||
if (totalSize > lowWatermarkBytes) {
|
||||
|
|
Loading…
Reference in a new issue